One document matched: draft-ietf-dccp-rfc3448bis-02.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.18.1
%%CreationDate: Sun Jul  8 20:46:11 2007
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.18 1
%%Pages: 39
%%PageOrder: Ascend
%%Orientation: Portrait
%%EndComments
%%BeginProlog
%%BeginResource: procset grops 1.18 1
/setpacking where{
pop
currentpacking
true setpacking
}if
/grops 120 dict dup begin
/SC 32 def
/A/show load def
/B{0 SC 3 -1 roll widthshow}bind def
/C{0 exch ashow}bind def
/D{0 exch 0 SC 5 2 roll awidthshow}bind def
/E{0 rmoveto show}bind def
/F{0 rmoveto 0 SC 3 -1 roll widthshow}bind def
/G{0 rmoveto 0 exch ashow}bind def
/H{0 rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/I{0 exch rmoveto show}bind def
/J{0 exch rmoveto 0 SC 3 -1 roll widthshow}bind def
/K{0 exch rmoveto 0 exch ashow}bind def
/L{0 exch rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/M{rmoveto show}bind def
/N{rmoveto 0 SC 3 -1 roll widthshow}bind def
/O{rmoveto 0 exch ashow}bind def
/P{rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/Q{moveto show}bind def
/R{moveto 0 SC 3 -1 roll widthshow}bind def
/S{moveto 0 exch ashow}bind def
/T{moveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/SF{
findfont exch
[exch dup 0 exch 0 exch neg 0 0]makefont
dup setfont
[exch/setfont cvx]cvx bind def
}bind def
/MF{
findfont
[5 2 roll
0 3 1 roll
neg 0 0]makefont
dup setfont
[exch/setfont cvx]cvx bind def
}bind def
/level0 0 def
/RES 0 def
/PL 0 def
/LS 0 def
/MANUAL{
statusdict begin/manualfeed true store end
}bind def
/PLG{
gsave newpath clippath pathbbox grestore
exch pop add exch pop
}bind def
/BP{
/level0 save def
1 setlinecap
1 setlinejoin
72 RES div dup scale
LS{
90 rotate
}{
0 PL translate
}ifelse
1 -1 scale
}bind def
/EP{
level0 restore
showpage
}bind def
/DA{
newpath arcn stroke
}bind def
/SN{
transform
.25 sub exch .25 sub exch
round .25 add exch round .25 add exch
itransform
}bind def
/DL{
SN
moveto
SN
lineto stroke
}bind def
/DC{
newpath 0 360 arc closepath
}bind def
/TM matrix def
/DE{
TM currentmatrix pop
translate scale newpath 0 0 .5 0 360 arc closepath
TM setmatrix
}bind def
/RC/rcurveto load def
/RL/rlineto load def
/ST/stroke load def
/MT/moveto load def
/CL/closepath load def
/Fr{
setrgbcolor fill
}bind def
/Fk{
setcmykcolor fill
}bind def
/Fg{
setgray fill
}bind def
/FL/fill load def
/LW/setlinewidth load def
/Cr/setrgbcolor load def
/Ck/setcmykcolor load def
/Cg/setgray load def
/RE{
findfont
dup maxlength 1 index/FontName known not{1 add}if dict begin
{
1 index/FID ne{def}{pop pop}ifelse
}forall
/Encoding exch def
dup/FontName exch def
currentdict end definefont pop
}bind def
/DEFS 0 def
/EBEGIN{
moveto
DEFS begin
}bind def
/EEND/end load def
/CNT 0 def
/level1 0 def
/PBEGIN{
/level1 save def
translate
div 3 1 roll div exch scale
neg exch neg exch translate
0 setgray
0 setlinecap
1 setlinewidth
0 setlinejoin
10 setmiterlimit
[]0 setdash
/setstrokeadjust where{
pop
false setstrokeadjust
}if
/setoverprint where{
pop
false setoverprint
}if
newpath
/CNT countdictstack def
userdict begin
/showpage{}def
}bind def
/PEND{
clear
countdictstack CNT sub{end}repeat
level1 restore
}bind def
end def
/setpacking where{
pop
setpacking
}if
%%EndResource
%%IncludeResource: font Courier-Bold
%%IncludeResource: font Times-Bold
%%IncludeResource: font Times-Roman
%%IncludeResource: font Courier
grops begin/DEFS 1 dict def DEFS begin/u{.001 mul}bind def end/RES 72
def/PL 841.89 def/LS false def/ENC0[/asciicircum/asciitilde/Scaron
/Zcaron/scaron/zcaron/Ydieresis/trademark/quotesingle/Euro/.notdef
/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef
/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef
/.notdef/.notdef/.notdef/space/exclam/quotedbl/numbersign/dollar/percent
/ampersand/quoteright/parenleft/parenright/asterisk/plus/comma/hyphen
/period/slash/zero/one/two/three/four/five/six/seven/eight/nine/colon
/semicolon/less/equal/greater/question/at/A/B/C/D/E/F/G/H/I/J/K/L/M/N/O
/P/Q/R/S/T/U/V/W/X/Y/Z/bracketleft/backslash/bracketright/circumflex
/underscore/quoteleft/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/x/y
/z/braceleft/bar/braceright/tilde/.notdef/quotesinglbase/guillemotleft
/guillemotright/bullet/florin/fraction/perthousand/dagger/daggerdbl
/endash/emdash/ff/fi/fl/ffi/ffl/dotlessi/dotlessj/grave/hungarumlaut
/dotaccent/breve/caron/ring/ogonek/quotedblleft/quotedblright/oe/lslash
/quotedblbase/OE/Lslash/.notdef/exclamdown/cent/sterling/currency/yen
/brokenbar/section/dieresis/copyright/ordfeminine/guilsinglleft
/logicalnot/minus/registered/macron/degree/plusminus/twosuperior
/threesuperior/acute/mu/paragraph/periodcentered/cedilla/onesuperior
/ordmasculine/guilsinglright/onequarter/onehalf/threequarters
/questiondown/Agrave/Aacute/Acircumflex/Atilde/Adieresis/Aring/AE
/Ccedilla/Egrave/Eacute/Ecircumflex/Edieresis/Igrave/Iacute/Icircumflex
/Idieresis/Eth/Ntilde/Ograve/Oacute/Ocircumflex/Otilde/Odieresis
/multiply/Oslash/Ugrave/Uacute/Ucircumflex/Udieresis/Yacute/Thorn
/germandbls/agrave/aacute/acircumflex/atilde/adieresis/aring/ae/ccedilla
/egrave/eacute/ecircumflex/edieresis/igrave/iacute/icircumflex/idieresis
/eth/ntilde/ograve/oacute/ocircumflex/otilde/odieresis/divide/oslash
/ugrave/uacute/ucircumflex/udieresis/yacute/thorn/ydieresis]def
/Courier@0 ENC0/Courier RE/Times-Roman@0 ENC0/Times-Roman RE
/Times-Bold@0 ENC0/Times-Bold RE/Courier-Bold@0 ENC0/Courier-Bold RE
%%EndProlog
%%Page: 1 1
%%BeginPageSetup
BP
%%EndPageSetup
/F0 10/Courier-Bold@0 SF(Internet Engineering Task Force)72 85 Q
(S. Floyd)197.998 E 317.998(INTERNET-DRAFT ICIR)72 98 R
(Intended status: Proposed Standard)72 111 Q(M. Handley)167.998 E
(Expires: January 2008)72 124 Q(University College London)155.998 E
(J. Padhye)449.998 137 Q(Microsoft)449.998 150 Q(J. Widmer)449.998 163 Q
(University of Mannheim)371.998 176 Q 6(8J)437.998 189 S(uly 2007)-6 E
/F1 14/Times-Bold@0 SF(TCP Friendly Rate Contr)111.327 214 Q
(ol \(TFRC\): Pr)-.252 E(otocol Speci\214cation)-.252 E
(draft-ietf-dccp-rfc3448bis-02.ps)193.325 227 Q/F2 11/Times-Bold@0 SF
(Status of this Memo)72 259 Q/F3 11/Times-Roman@0 SF
(By submitting this Internet-Draft, each author represents that an)100.8
275.6 Q 2.75(ya)-.165 G(pplicable patent or other)-2.75 E
(IPR claims of which he or she is a)100.8 288.6 Q -.11(wa)-.165 G(re ha)
.11 E .33 -.165(ve b)-.22 H(een or will be disclosed, and an).165 E 2.75
(yo)-.165 G 2.75(fw)-2.75 G(hich he)-2.75 E(or she becomes a)100.8 301.6
Q -.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)100.8 318.2 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)100.8 331.2 Q
(orking groups.)-.11 E(Note that other groups may also distrib)5.5 E
(ute w)-.22 E(orking)-.11 E(documents as Internet-Drafts.)100.8 344.2 Q
(Internet-Drafts are draft documents v)100.8 360.8 Q
(alid for a maximum of six months and may be)-.275 E
(updated, replaced, or obsoleted by other documents at an)100.8 373.8 Q
2.75(yt)-.165 G 2.75(ime. It)-2.75 F(is inappropriate to use)2.75 E
(Internet-Drafts as reference material or to cite them other than as "w)
100.8 386.8 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)100.8
403.4 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)100.8 416.4
Q(The list of Internet-Draft Shado)100.8 433 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)100.8 446 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)100.8 462.6 Q(xpire on January 2008.)-.165 E
F2(Copyright Notice)72 488.6 Q F3(Cop)100.8 505.2 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E F2(Abstract)72
531.2 Q F3
(This document speci\214es TCP-Friendly Rate Control \(TFRC\).)100.8
547.8 Q(TFRC is a congestion)5.5 E(control mechanism for unicast \215o)
100.8 560.8 Q(ws operating in a best-ef)-.275 E(fort Internet en)-.275 E
2.75(vironment. It)-.44 F(is)2.75 E(reasonably f)100.8 573.8 Q
(air when competing for bandwidth with TCP \215o)-.11 E(ws, b)-.275 E
(ut has a much lo)-.22 E(wer)-.275 E -.275(va)100.8 586.8 S
(riation of throughput o).275 E -.165(ve)-.165 G 2.75(rt).165 G
(ime compared with TCP)-2.75 E 2.75(,m)-1.221 G
(aking it more suitable for)-2.75 E
(applications such as streaming media where a relati)100.8 599.8 Q -.165
(ve)-.275 G(ly smooth sending rate is of).165 E(importance.)100.8 612.8
Q(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
252.843(idmer [P)-.44 F(age 1])-.165 E 0 Cg EP
%%Page: 2 2
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E/F1 11/Times-Bold@0 SF -1.012
(Ta)72 85 S(ble of Contents)1.012 E/F2 10/Times-Roman@0 SF
(1. Introduction)100.8 110 Q F0 -2.25(.................................\
.......................................................................\
.............. ....)2.76 F F2(8)A(2. Con)100.8 122 Q -.15(ve)-.4 G
(ntions).15 E F0 -2.25(................................................\
..................................................................... .\
...)4.94 F F2(9)A(3. Protocol Mechanism)100.8 134 Q F0 -2.25(..........\
.......................................................................\
........................ ....)4.9 F F2(9)A(3.1. TCP Throughput Equation)
115.8 146 Q F0 -2.25(..................................................\
....................................... ....)2.78 F F2(9)A(3.2. P)115.8
158 Q(ack)-.15 E(et Contents)-.1 E F0 -1.75(...........................\
.......................................................................\
...... ..)3.73 F F2(10)A(3.2.1. Data P)140.8 170 Q(ack)-.15 E(ets)-.1 E
F0 -1.75(..............................................................\
................................... ..)3.27 F F2(11)A(3.2.2. Feedback P)
140.8 182 Q(ack)-.15 E(ets)-.1 E F0 -1.75(.............................\
............................................................. ..)3.08 F
F2(11)A(4. Data Sender Protocol)100.8 194 Q F0 -1.75(..................\
.......................................................................\
............... ..)5.16 F F2(12)A(4.1. Measuring the Se)115.8 206 Q
(gment Size)-.15 E F0 -1.75(...........................................\
........................................... ..)3.14 F F2(12)A
(4.2. Sender Initialization)115.8 218 Q F0 -1.75(......................\
.......................................................................\
..... ..)3.32 F F2(13)A(4.3. Sender Beha)115.8 230 Q
(vior When a Feedback P)-.2 E(ack)-.15 E(et is Recei)-.1 E -.15(ve)-.25
G(d).15 E F0 -1.75(................................................ ..)
3.94 F F2(13)A(4.4. Expiration of Nofeedback T)115.8 242 Q(imer)-.35 E
F0 -1.75(..............................................................\
.................. ..)3.75 F F2(15)A(4.5. Reducing Oscillations)115.8
254 Q F0 -1.75(........................................................\
....................................... ..)4.34 F F2(16)A
(4.6. Scheduling of P)115.8 266 Q(ack)-.15 E(et T)-.1 E(ransmissions)
-.35 E F0 -1.75(.......................................................\
.................... ..)3.28 F F2(17)A(4.6.1. Sending P)140.8 278 Q(ack)
-.15 E(ets Before their Nominal Send T)-.1 E(ime)-.35 E F0 -1.75
(........................................... ..)4.07 F F2(18)A
(5. Calculation of the Loss Ev)100.8 290 Q(ent Rate \(p\))-.15 E F0
-1.75(.................................................................\
............... ..)2.98 F F2(19)A(5.1. Detection of Lost or Mark)115.8
302 Q(ed P)-.1 E(ack)-.15 E(ets)-.1 E F0 -1.75(........................\
.................................................. ..)3.58 F F2(19)A
(5.2. T)115.8 314 Q(ranslation from Loss History to Loss Ev)-.35 E(ents)
-.15 E F0 -1.75
(............................................................. ..)4.18 F
F2(20)A(5.3. Inter)115.8 326 Q(-loss Ev)-.2 E(ent Interv)-.15 E(al)-.25
E F0 -1.75(............................................................\
................................ ..)4.6 F F2(21)A(5.4. A)115.8 338 Q
-.15(ve)-.74 G(rage Loss Interv).15 E(al)-.25 E F0 -1.75(..............\
.......................................................................\
.......... ..)5.23 F F2(21)A(5.5. History Discounting)115.8 350 Q F0
-1.75(.................................................................\
................................. ..)3.31 F F2(22)A(6. Data Recei)100.8
362 Q -.15(ve)-.25 G 2.5(rP).15 G(rotocol)-2.5 E F0 -1.75(.............\
.......................................................................\
.................. ..)3.29 F F2(24)A(6.1. Recei)115.8 374 Q -.15(ve)-.25
G 2.5(rB).15 G(eha)-2.5 E(vior When a Data P)-.2 E(ack)-.15 E
(et is Recei)-.1 E -.15(ve)-.25 G(d).15 E F0 -1.75
(.................................................... ..)5.01 F F2(24)A
(6.2. Expiration of Feedback T)115.8 386 Q(imer)-.35 E F0 -1.75(.......\
.......................................................................\
...... ..)2.74 F F2(25)A(6.3. Recei)115.8 398 Q -.15(ve)-.25 G 2.5(rI)
.15 G(nitialization)-2.5 E F0 -1.75(...................................\
............................................................ ..)4.2 F F2
(25)A(6.3.1. Initializing the Loss History after the First Loss Ev)140.8
410 Q(ent)-.15 E F0 -1.75(..................................... ..)4 F
F2(26)A(7. Sender)100.8 422 Q(-based V)-.2 E(ariants)-1.11 E F0 -1.75(.\
.......................................................................\
............................... ..)4.51 F F2(27)A
(8. Implementation Issues)100.8 434 Q F0 -1.75(........................\
.......................................................................\
........ ..)3.18 F F2(27)A(9. Changes from RFC 3448)100.8 446 Q F0 -1.75
(......................................................................\
............................. ..)4.17 F F2(28)A
(10. Security Considerations)100.8 458 Q F0 -1.75(.....................\
.......................................................................\
....... ..)3.62 F F2(30)A(11. IAN)100.8 470 Q 2.5(AC)-.35 G
(onsiderations)-2.5 E F0 -1.75(........................................\
.............................................................. ..)4.06 F
F2(31)A(12. Ackno)100.8 482 Q(wledgments)-.25 E F0 -1.75(..............\
.......................................................................\
...................... ..)2.71 F F2(31)A(A. T)100.8 494 Q(erminology)-.7
E F0 -1.75(............................................................\
........................................................ ..)3.96 F F2
(31)A(B. The Initial V)100.8 506 Q(alue of the Nofeedback T)-1.11 E
(imer)-.35 E F0 -1.75(.................................................\
........................ ..)4.12 F F2(33)A
(C. Response to Idle or Data-limited Periods)100.8 518 Q F0 -1.75(.....\
.......................................................................\
 ..)2.99 F F2(33)A(C.1. Long Idle or Data-limited Periods)115.8 530 Q F0
-1.75(.................................................................\
............ ..)5.24 F F2(33)A(C.2. Short Idle or Data-limited Periods)
115.8 542 Q F0 -1.75(..................................................\
........................... ..)4.68 F F2(35)A
(C.3. Moderate Idle or Data-limited Periods)115.8 554 Q F0 -1.75(......\
................................................................. ..)
4.53 F F2(36)A(C.4. Other P)115.8 566 Q(atterns)-.15 E F0 -1.75(.......\
.......................................................................\
............................ ..)3.69 F F2(37)A(Normati)100.8 578 Q .3
-.15(ve R)-.25 H(eferences).15 E F0 -1.75(.............................\
.......................................................................\
....... ..)3.16 F F2(37)A(Informational References)100.8 590 Q F0 -1.75
(......................................................................\
................................ ..)4.29 F F2(37)A(Authors' Addresses)
100.8 602 Q F0 -1.75(..................................................\
............................................................ ..)5.05 F
F2(38)A(Full Cop)100.8 614 Q(yright Statement)-.1 E F0 -1.75(..........\
.......................................................................\
..................... ..)4.62 F F2(38)A(Intellectual Property)100.8 626
Q F0 -1.75(............................................................\
................................................. ..)5.02 F F2(39)A F0
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
252.843(idmer [P)-.44 F(age 2])-.165 E 0 Cg EP
%%Page: 3 3
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(NO)100.8 85 Q(TE T)-.44 E 2.75
(OR)-.198 G(FC EDIT)-2.75 E(OR: PLEASE DELETE THIS NO)-.198 E
(TE UPON PUBLICA)-.44 E(TION.)-1.221 E
(Changes from draft-ietf-dccp-rfc3448bis-01.txt:)103.55 111 Q 2.75(*S)
103.55 137 S(peci\214ed that the sender is not limited by the recei)
-2.75 E .33 -.165(ve r)-.275 H(ate).165 E
(if the sender has been data-limited for an entire feedback)109.05 150 Q
(interv)109.05 163 Q(al.)-.275 E 2.75(*A)103.55 189 S(dded v)-2.75 E
(ariables "initial_rate" and "reco)-.275 E -.165(ve)-.165 G
(r_rate, for the).165 E
(initial transmit rate and the rate for resuming after an idle)109.05
202 Q(period, for easier speci\214cation of F)109.05 215 Q
(aster Restart \(in a separate)-.165 E 2.75(document\). Also)109.05 228
R(added the v)2.75 E(ariable "recv_limit" to specify)-.275 E
(the limit on the sending rate that is computed from the recei)109.05
241 Q -.165(ve)-.275 G(rate, and the v)109.05 254 Q
(ariable "timer_limit" to specify the)-.275 E
(limit on the sending rate from the e)109.05 267 Q
(xpiration of the nofeedback)-.165 E(timer)109.05 280 Q(.)-.605 E
(Explained wh)109.05 293 Q 2.75(yr)-.055 G(eco)-2.75 E -.165(ve)-.165 G
(r_rate is not used as lo).165 E(wer bound)-.275 E
(for nofeedback timer e)109.05 306 Q
(xpirations after a data-limited period.)-.165 E 2.75(*A)103.55 332 S
(dded Appendix C on "Response to Idle or Data-limited Periods".)-2.75 E
2.75(*R)103.55 358 S -.275(ev)-2.75 G
(ised the section on "Scheduling of P).275 E(ack)-.165 E(et T)-.11 E
(ransmissions")-.385 E(to mak)109.05 371 Q 2.75(ec)-.11 G
(lear what is speci\214cation, and what is)-2.75 E 2.75
(implementation. From)109.05 384 R 2.75(Gerrit. Also)2.75 F
(stated that the)2.75 E
(accumulation of sending credits should be limited)109.05 397 Q
(to a round-trip time')109.05 410 Q 2.75(sw)-.605 G(orth of pack)-2.86 E
(ets.)-.11 E 2.75(*F)103.55 436 S(or measuring the recei)-2.915 E .33
-.165(ve r)-.275 H(ate, added that after a loss e).165 E -.165(ve)-.275
G(nt,).165 E(the recei)109.05 449 Q .33 -.165(ve r)-.275 H
(ate SHOULD be measured o).165 E -.165(ve)-.165 G 2.75(rt).165 G
(he most recent R)-2.75 E(TT)-.66 E(,)-.814 E -.22(bu)109.05 462 S 2.75
(tf).22 G(or simplicity of implementation, MA)-2.75 E 2.75(Yb)-1.155 G
2.75(em)-2.75 G(easured o)-2.75 E -.165(ve)-.165 G(r).165 E 2.75(as)
109.05 475 S(lightly longer time interv)-2.75 E(al.)-.275 E 2.75(*C)
103.55 501 S(lari\214ed that R)-2.75 E(TT measurements don')-.66 E 2.75
(tn)-.198 G(ecessarily come from)-2.75 E(feedback pack)109.05 514 Q
(ets; the)-.11 E 2.75(yc)-.165 G(ould also come from other places,)-2.75
E(e.g., from the SYN e)109.05 527 Q(xchange.)-.165 E 2.75(*S)103.55 553
S(peci\214ed that the sender may maintain unused sent credits)-2.75 E
(up to one R)109.05 566 Q(TT)-.66 E 5.5(.T)-.814 G(his gi)-5.5 E -.165
(ve)-.275 G 2.75(sb).165 G(eha)-2.75 E(vior similar to TCP)-.22 E(.)
-1.221 E(Also speci\214ed that the sender should not sent pack)109.05
579 Q(ets more)-.11 E
(that rtt/2 seconds before their nominal send time.)109.05 592 Q 2.75
(*R)103.55 618 S
(einserted the last paragraph of Section 4.4 from RFC 3448.)-2.75 E
(It must ha)109.05 631 Q .33 -.165(ve b)-.22 H(een deleted accidently)
.165 E(.)-.715 E 2.75(*T)106.3 657 S(ODO in ns-2)-2.948 E 2.75(-A)111.8
670 S(dd a v)-2.75 E(ariable to ns-2 to allo)-.275 E 2.75(we)-.275 G
(ither TFRC or CCID3.)-2.75 E 2.75(*F)103.55 696 S
(eedback from Arjuna Sathiaseelan:)-2.75 E(Handle)72 733 Q(y/Flo)-.165 E
(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 252.843(idmer [P)-.44 F(age 3])
-.165 E 0 Cg EP
%%Page: 4 4
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 2.75(-C)109.05 85 S
(hanging W_init to be in terms of se)-2.75 E(gment size s, not MSS.)
-.165 E 2.75(*C)103.55 111 S(hanged THRESHOLD, the lo)-2.75 E
(wer bound on the history)-.275 E(discounting parameter DF)109.05 124 Q
2.75(,f)-.88 G(rom 0.5 to 0.25, for more)-2.75 E
(history discounting when the current interv)109.05 137 Q(al is long.)
-.275 E 2.75(*R)103.55 163 S
(elying on the sender not to use X_recv from data-limited)-2.75 E 2.75
(periods. This)109.05 176 R(gi)2.75 E -.165(ve)-.275 G 2.75(sb).165 G
(eha)-2.75 E(vior similar to TCP)-.22 E 2.75(,w)-1.221 G(hen)-2.75 E
-.44(AC)109.05 189 S(K-clocking is not in ef).44 E
(fect in data-limited periods.)-.275 E(The lar)109.05 202 Q
(gest X_recv o)-.198 E -.165(ve)-.165 G 2.75(rt).165 G
(he most recent tw)-2.75 E 2.75(or)-.11 G(ound-trip)-2.75 E
(times is used to limit the sending rate.)109.05 215 Q(This is)5.5 E
(maintained using X_recv_set.)109.05 228 Q -.88(Ta)5.5 G -.11(ke).88 G
2.75(nt).11 G(ogether)-2.75 E 2.75(,t)-.44 G(hese a)-2.75 E -.22(vo)-.22
G(id).22 E(problems with the \214rst feedback pack)109.05 241 Q
(et after an idle)-.11 E(period, and this a)109.05 254 Q -.22(vo)-.22 G
(ids problems with limitations).22 E
(from X_recv during data-limited periods.)109.05 267 Q 2.75(*C)103.55
293 S(lari\214ed that when the recei)-2.75 E -.165(ve)-.275 G 2.75(rr)
.165 G(ecei)-2.75 E -.165(ve)-.275 G 2.75(sad).165 G(ata pack)-2.75 E
(et,)-.11 E(and didn')109.05 306 Q 2.75(ts)-.198 G(end a feedback pack)
-2.75 E(et when the feedback timer)-.11 E(last e)109.05 319 Q
(xpired \(because no data pack)-.165 E(ets were recei)-.11 E -.165(ve)
-.275 G(d\),).165 E(then the recei)109.05 332 Q -.165(ve)-.275 G 2.75
(rs).165 G(ends a feedback pack)-2.75 E(et immediately)-.11 E(.)-.715 E
2.75(*C)103.55 358 S(lari\214ed that the feedback pack)-2.75 E
(et reports the rate o)-.11 E -.165(ve)-.165 G(r).165 E(the last R)
109.05 371 Q(TT)-.66 E 2.75(,n)-.814 G
(ot necessarily the rate since the)-2.75 E(last feedback pack)109.05 384
Q(et w)-.11 E(as sent \(if no feedback pack)-.11 E(et w)-.11 E(as)-.11 E
(sent when the feedback timer last e)109.05 397 Q(xpired\).)-.165 E 2.75
(*C)103.55 423 S(orrected earlier code designed to pre)-2.75 E -.165(ve)
-.275 G(nt the recei).165 E -.165(ve)-.275 G
(rate from limiting the sending rate when the \214rst feedback)109.05
436 Q(pack)109.05 449 Q(et recei)-.11 E -.165(ve)-.275 G
(d, or for the \214rst feedback pack).165 E(et recei)-.11 E -.165(ve)
-.275 G(d).165 E(after an idle period.)109.05 462 Q 2.75(*C)103.55 488 S
(lari\214ed that we ha)-2.75 E .33 -.165(ve p)-.22 H
(=0 only until the \214rst loss e).165 E -.165(ve)-.275 G(nt.).165 E
(After the \214rst loss e)109.05 501 Q -.165(ve)-.275 G
(nt, p>0, and it is not possible to go).165 E(back to p=0.)109.05 514 Q
(In response to old email.)5.5 E 2.75(*C)103.55 540 S
(lari\214ed in Section 6.1 that the loss e)-2.75 E -.165(ve)-.275 G
(nt rate does not).165 E(ha)109.05 553 Q .33 -.165(ve t)-.22 H 2.75(ob)
.165 G 2.75(er)-2.75 G(ecalculated with the arri)-2.75 E -.275(va)-.275
G 2.75(lo).275 G 2.75(fe)-2.75 G(ach ne)-2.75 E 2.75(wd)-.275 G(ata)
-2.75 E(pack)109.05 566 Q(et.)-.11 E 2.75(*C)103.55 592 S
(lari\214ed the section on Reducing Oscillations.)-2.75 E(Feedback from)
5.5 E(Gerrit Renk)109.05 605 Q(er)-.11 E(.)-.605 E
(Changes from draft-ietf-dccp-rfc3448bis-00.txt:)103.55 631 Q 2.75(*W)
103.55 657 S(hen initializing the loss history after the \214rst)-2.75 E
(data pack)109.05 670 Q(et sent is lost or ECN-mark)-.11 E
(ed, TFRC uses)-.11 E 2.75(am)109.05 683 S(inimum recei)-2.75 E .33
-.165(ve r)-.275 H(ate of 0.5 pack).165 E(ets per second.)-.11 E(Handle)
72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 252.843
(idmer [P)-.44 F(age 4])-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 2.75(*F)103.55 85 S
(or initializing the estimated pack)-2.915 E(et drop rate)-.11 E
(for the \214rst loss interv)109.05 98 Q(al when coming out of slo)-.275
E(w-start,)-.275 E(it is ok to use the maximum recei)109.05 111 Q .33
-.165(ve r)-.275 H(ate so f).165 E(ar)-.11 E 2.75(,n)-.44 G(ot just)
-2.75 E(the recei)109.05 124 Q .33 -.165(ve r)-.275 H
(ate in the last round-trip time.).165 E(Feedback from Ladan Gharai.)
109.05 137 Q 2.75(*G)103.55 163 S(eneral feedback from Gorry F)-2.75 E
(airhurst:)-.165 E 2.75(-A)109.05 176 S(dded a reference for RFC4828.)
-2.75 E 2.75(-C)109.05 189 S(lari\214ed that R_m is sender')-2.75 E 2.75
(se)-.605 G(stimate of R)-2.75 E(TT)-.66 E 2.75(,a)-.814 G 2.75(sr)-2.75
G(eported)-2.75 E(in Section 3.2.1.)114.55 202 Q 2.75(-A)109.05 215 S
(dded a de\214nition of terms.)-2.75 E 2.75(-A)109.05 228 S
(dded a discussion of wh)-2.75 E 2.75(yt)-.055 G(he initial v)-2.75 E
(alue of the nofeedback)-.275 E(timer is tw)114.55 241 Q 2.75(os)-.11 G
(econds, instead of three seconds for the)-2.75 E(recommended initial v)
114.55 254 Q(alue for TCP')-.275 E 2.75(sr)-.605 G(etransmit timer)-2.75
E(.)-.605 E 2.75(*G)103.55 280 S
(eneral feedback from Arjuna Sathiaseelan:)-2.75 E 2.75(-A)109.05 293 S
(dded more details about sending multiple feedback)-2.75 E(pack)117.3
306 Q(ets per R)-.11 E(TT)-.66 E(.)-.814 E 2.75(-A)109.05 319 S
(dded change to Section 4.3 to use the \214rst feedback)-2.75 E(pack)
117.3 332 Q(et, or the \214rst feedback pack)-.11 E(et after a)-.11 E
(nofeedback timer during slo)117.3 345 Q(w-start, *if min_rate > X*.)
-.275 E 2.75(*G)103.55 371 S(eneral feedback from Gerrit Renk)-2.75 E
(er:)-.11 E 2.75(-C)109.05 384 S(hanged "delta" to "t_delta".)-2.75 E
2.75(-C)109.05 397 S(hanged X_calc to X_Bps, clari\214ed X.)-2.75 E 2.75
(-C)109.05 410 S(lari\214ed send times in "Scheduling of P)-2.75 E(ack)
-.165 E(et T)-.11 E(ransmissions".)-.385 E 2.75(-C)109.05 423 S
(hanged so that tld can be initialized to either 0 or -1.)-2.75 E 2.75
(-F)109.05 436 S(ix)-2.75 E
(ed Section 5.5 to say that the most recent lost)-.165 E(interv)114.55
449 Q(al has weight 1/\(0.75*n\) *when there ha)-.275 E .33 -.165(ve b)
-.22 H(een).165 E(at least eight loss interv)114.55 462 Q(als*.)-.275 E
2.75(-C)109.05 475 S(lari\214ed introduction about \214x)-2.75 E
(ed-size and v)-.165 E(ariable-size)-.275 E(pack)114.55 488 Q(ets.)-.11
E 2.75(*A)103.55 514 S(dded more about sender)-2.75 E(-based v)-.22 E
(ariants.)-.275 E(Feedback from Guillaume Jourjon.)109.05 527 Q 2.75(*C)
103.55 553 S(orrected that the loss interv)-2.75 E
(al I_0 includes all transmitted)-.275 E(pack)109.05 566 Q
(ets, including lost and mark)-.11 E(ed pack)-.11 E
(ets \(as de\214ned in Section)-.11 E
(5.3 in the general de\214nition.\))109.05 579 Q(Email from Eddie K)5.5
E(ohler and)-.385 E(Gerrit Renk)109.05 592 Q(er)-.11 E(.)-.605 E 2.75
(*N)103.55 618 S(ot done:)-2.75 E 2.75(Id)5.5 G(idn')-2.75 E 2.75(ta)
-.198 G(dd a minimum v)-2.75 E(alue for the nofeedback)-.275 E(timer)
109.05 631 Q 5.5(.\()-.605 G(Wh)-5.5 E 2.75(yw)-.055 G
(ould a nofeedback timer need to be bigger)-2.86 E
(than max\(4*R, 2*s/X\)?)109.05 644 Q
(Email discussing pros and cons from)5.5 E(Arjuna.)109.05 657 Q
(Changes from draft-\215o)103.55 683 Q(yd-rfc3448bis-00.txt:)-.11 E
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
252.843(idmer [P)-.44 F(age 5])-.165 E 0 Cg EP
%%Page: 6 6
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 2.75(*N)103.55 85 S
(ame change to draft-ietf-dccp-rfc3448bis-00.txt.)-2.75 E 2.75(*S)103.55
111 S(peci\214ed the recei)-2.75 E -.165(ve)-.275 G(r').165 E 2.75(si)
-.605 G(nitialization of the feedback timer)-2.75 E
(when the \214rst data pack)109.05 124 Q(et doesn')-.11 E 2.75(th)-.198
G -2.475 -.22(av e)-2.75 H(an estimate of the)2.97 E -.66(RT)109.05 137
S 4.378 -.814(T. F).66 H(rom feedback from Dado Colussi.).814 E 2.75(*A)
103.55 163 S(dded the procedure for sending recei)-2.75 E -.165(ve)-.275
G(r).165 E(feedback pack)109.05 176 Q(ets when a coarse-grained)-.11 E
(timestamp is used. From RFC 4243.)109.05 189 Q(Changes from RFC 3448:)
103.55 215 Q 2.75(*I)103.55 241 S
(ncorporated changes in the RFC 3448 errata:)-2.75 E 5.5(-")109.05 267 S
(If the sender does not recei)-5.5 E .33 -.165(ve a f)-.275 H
(eedback report for).165 E
(four round trip times, it cuts its sending rate in half.")117.3 280 Q
(\("T)117.3 293 Q -.11(wo)-.88 G 2.75("c).11 G
(hanged to "four", for consistenc)-2.75 E 2.75(yw)-.165 G(ith the rest)
-2.75 E(of the document.)117.3 306 Q(Reported by Joer)5.5 E 2.75(gW)
-.198 G(idmer\).)-3.19 E 2.75(-")109.05 332 S(If the nofeedback timer e)
-2.75 E(xpires when the sender does not)-.165 E(yet ha)114.55 345 Q .33
-.165(ve a)-.22 H 2.75(nR).165 G(TT sample, and has not yet recei)-3.41
E -.165(ve)-.275 G 2.75(da).165 G -.165(ny)-2.75 G
(feedback from the recei)114.55 358 Q -.165(ve)-.275 G .88 -.44(r, o)
.165 H 2.75(rw).44 G(hen p == 0,...")-2.75 E
(\(Added "or when p == 0,", reported by W)114.55 371 Q(im Heirman\).)
-.44 E 2.75(-I)109.05 397 S 2.75(nS)-2.75 G(ection 5.5, changed:)-2.75 E
(for \(i = 1 to n\) { DF_i = 1; })120.05 410 Q(to:)114.55 423 Q
(for \(i = 0 to n\) { DF_i = 1; })120.05 436 Q(Reported by Michele R.)
114.55 449 Q 2.75(*C)103.55 475 S
(hanged RFC 3448 to correspond to the lar)-2.75 E(ger initial windo)
-.198 E(ws)-.275 E(speci\214ed in RFC 3390.)109.05 488 Q
(This includes the follo)5.5 E(wing:)-.275 E 2.75(-I)109.05 514 S
(ncorporated Section 5.1 from [RFC4342], saying that)-2.75 E
(when reducing the sending rate after an idle period, don')114.55 527 Q
(t)-.198 E(reduce the sending rate belo)114.55 540 Q 2.75(wt)-.275 G
(he initial sending rate.)-2.75 E 2.75(-C)109.05 566 S
(hange for a datalimited sender:)-2.75 E
(When the sender has been datalimited, the sender doesn')114.55 579 Q(t)
-.198 E(let the recei)114.55 592 Q .33 -.165(ve r)-.275 H
(ate limit it to a sending rate less than).165 E(the initial rate.)
114.55 605 Q 2.75(-S)109.05 631 S(mall change to slo)-2.75 E(w-start:)
-.275 E(Changed so that for the \214rst feedback pack)114.55 644 Q
(et recei)-.11 E -.165(ve)-.275 G(d,).165 E
(or for the \214rst feedback pack)114.55 657 Q(et recei)-.11 E -.165(ve)
-.275 G 2.75(da).165 G(fter an idle)-2.75 E(period, the recei)114.55 670
Q .33 -.165(ve r)-.275 H(ate is not used to limit the).165 E
(sending rate.)114.55 683 Q(This is because the recei)5.5 E -.165(ve)
-.275 G 2.75(rm).165 G(ight not yet)-2.75 E(ha)114.55 696 Q .33 -.165
(ve s)-.22 H(een an entire windo).165 E 2.75(wo)-.275 G 2.75(fd)-2.75 G
(ata.)-2.75 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E
(ye/W)-.055 E 252.843(idmer [P)-.44 F(age 6])-.165 E 0 Cg EP
%%Page: 7 7
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 2.75(*C)103.55 85 S
(lari\214ed ho)-2.75 E 2.75(wt)-.275 G(he a)-2.75 E -.165(ve)-.22 G
(rage loss interv).165 E(al is calculated when)-.275 E(the recei)109.05
98 Q -.165(ve)-.275 G 2.75(rh).165 G(as not yet seen eight loss interv)
-2.75 E(als.)-.275 E 2.75(*D)103.55 124 S
(iscussed more about estimating the a)-2.75 E -.165(ve)-.22 G(rage se)
.165 E(gment size:)-.165 E 2.75(-F)109.05 150 S
(or initializing the loss history after the \214rst loss e)-2.915 E
-.165(ve)-.275 G(nt,).165 E(either the recei)114.55 163 Q -.165(ve)-.275
G 2.75(rk).165 G(no)-2.75 E(ws the sender')-.275 E 2.75(sv)-.605 G
(alue for s, or)-3.025 E(the recei)114.55 176 Q -.165(ve)-.275 G 2.75
(ru).165 G(ses the throughput equation for X_pps and does)-2.75 E
(not need to kno)114.55 189 Q 2.75(wa)-.275 G 2.75(ne)-2.75 G
(stimate for s.)-2.75 E 2.75(-A)109.05 215 S
(dded a discussion about estimating the a)-2.75 E -.165(ve)-.22 G
(rage se).165 E(gment size)-.165 E 2.75(si)114.55 228 S 2.75(nS)-2.75 G
(ection 4.1 on "Measuring the Se)-2.75 E(gment Size".)-.165 E 2.75(-C)
109.05 254 S(hanged "pack)-2.75 E(et size" to "se)-.11 E(gment size".)
-.165 E(END OF NO)100.8 280 Q(TE T)-.44 E 2.75(OR)-.198 G(FC EDIT)-2.75
E(OR.)-.198 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E
(ye/W)-.055 E 252.843(idmer [P)-.44 F(age 7])-.165 E 0 Cg EP
%%Page: 8 8
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E/F1 11/Times-Bold@0 SF(1.)72 85
Q/F2 14/Times-Bold@0 SF(Intr)5.5 E(oduction)-.252 E F0
(This document speci\214es TCP-Friendly Rate Control \(TFRC\).)100.8
101.6 Q(TFRC is a congestion)5.5 E
(control mechanism designed for unicast \215o)100.8 114.6 Q
(ws operating in an Internet en)-.275 E(vironment and)-.44 E
(competing with TCP traf)100.8 127.6 Q
(\214c [FHPW00]. Instead of specifying a complete protocol, this)-.275 E
(document simply speci\214es a congestion control mechanism that could \
be used in a)100.8 140.6 Q(transport protocol such as DCCP \(Datagram C\
ongestion Control Protocol\) [RFC4340], in)100.8 153.6 Q(an application\
 incorporating end-to-end congestion control at the application le)100.8
166.6 Q -.165(ve)-.275 G(l, or in the).165 E(conte)100.8 179.6 Q(xt of \
endpoint congestion management [BRS99]. This document does not discuss)
-.165 E(pack)100.8 192.6 Q(et formats or reliability)-.11 E 5.5(.I)-.715
G(mplementation-related issues are discussed only brie\215y)-5.5 E 2.75
(,i)-.715 G(n)-2.75 E(Section 8.)100.8 205.6 Q
(TFRC is designed to be reasonably f)100.8 222.2 Q
(air when competing for bandwidth with TCP \215o)-.11 E(ws,)-.275 E
(where a \215o)100.8 235.2 Q 2.75(wi)-.275 G 2.75(s")-2.75 G
(reasonably f)-2.75 E(air" if its sending rate is generally within a f)
-.11 E(actor of tw)-.11 E 2.75(oo)-.11 G 2.75(ft)-2.75 G(he)-2.75 E
(sending rate of a TCP \215o)100.8 248.2 Q 2.75(wu)-.275 G
(nder the same conditions.)-2.75 E(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G
.88 -.44(r, T).165 H(FRC has a much lo).44 E(wer)-.275 E -.275(va)100.8
261.2 S(riation of throughput o).275 E -.165(ve)-.165 G 2.75(rt).165 G
(ime compared with TCP)-2.75 E 2.75(,w)-1.221 G(hich mak)-2.75 E
(es it more suitable for)-.11 E(applications such as telephon)100.8
274.2 Q 2.75(yo)-.165 G 2.75(rs)-2.75 G(treaming media where a relati)
-2.75 E -.165(ve)-.275 G(ly smooth sending rate).165 E
(is of importance.)100.8 287.2 Q(The penalty of ha)100.8 303.8 Q
(ving smoother throughput than TCP while competing f)-.22 E(airly for)
-.11 E(bandwidth is that TFRC responds slo)100.8 316.8 Q
(wer than TCP to changes in a)-.275 E -.275(va)-.22 G(ilable bandwidth.)
.275 E(Thus TFRC should only be used when the application has a require\
ment for smooth)100.8 329.8 Q(throughput, in particular)100.8 342.8 Q
2.75(,a)-.44 G -.22(vo)-2.97 G(iding TCP').22 E 2.75(sh)-.605 G
(alving of the sending rate in response to a single)-2.75 E(pack)100.8
355.8 Q(et drop.)-.11 E -.165(Fo)5.5 G 2.75(ra).165 G(pplications that \
simply need to transfer as much data as possible in as)-2.75 E
(short a time as possible we recommend using TCP)100.8 368.8 Q 2.75(,o)
-1.221 G 2.75(ri)-2.75 G 2.75(fr)-2.75 G
(eliability is not required, using an)-2.75 E(Additi)100.8 381.8 Q -.165
(ve)-.275 G(-Increase, Multiplicati).165 E -.165(ve)-.275 G
(-Decrease \(AIMD\) congestion control scheme with).165 E
(similar parameters to those used by TCP)100.8 394.8 Q(.)-1.221 E(TFRC \
is designed for best performance with applications that use a \214x)
100.8 411.4 Q(ed se)-.165 E(gment size, and)-.165 E -.275(va)100.8 424.4
S(ry their sending rate in pack).275 E
(ets per second in response to congestion.)-.11 E(TFRC can also be)5.5 E
(used, perhaps with less optimal performance, with applications that do\
n')100.8 437.4 Q 2.75(th)-.198 G -2.475 -.22(av e)-2.75 H 2.75<618c>2.97
G -.165(xe)-2.75 G(d).165 E(se)100.8 450.4 Q(gment size, b)-.165 E
(ut where the se)-.22 E(gment size v)-.165 E
(aries according to the needs of the application)-.275 E
(\(e.g., video applications\).)100.8 463.4 Q
(Some applications \(e.g., some audio applications\) require a \214x)
100.8 480 Q(ed interv)-.165 E(al of time between)-.275 E(pack)100.8 493
Q(ets and v)-.11 E(ary their se)-.275 E
(gment size instead of their pack)-.165 E
(et rate in response to congestion.)-.11 E(The congestion control mecha\
nism in this document is not designed for those applications;)100.8 506
Q(TFRC-SP \(Small-P)100.8 519 Q(ack)-.165 E(et TFRC\) is a v)-.11 E
(ariant of TFRC for applications that ha)-.275 E .33 -.165(ve a \214)
-.22 H -.165(xe).165 G(d).165 E(sending rate in pack)100.8 532 Q
(ets per second b)-.11 E(ut either use small pack)-.22 E(ets, or v)-.11
E(ary their pack)-.275 E(et size in)-.11 E(response to congestion.)100.8
545 Q(TFRC-SP is speci\214ed in a separate document [RFC4828].)5.5 E
(This document speci\214es TFRC as a recei)100.8 561.6 Q -.165(ve)-.275
G -.22(r-).165 G(based mechanism, with the calculation of the).22 E
(congestion control information \(i.e., the loss e)100.8 574.6 Q -.165
(ve)-.275 G(nt rate\) in the data recei).165 E -.165(ve)-.275 G 2.75(rr)
.165 G(ather in the)-2.75 E(data sender)100.8 587.6 Q 5.5(.T)-.605 G
(his is well-suited to an application where the sender is a lar)-5.5 E
(ge serv)-.198 E(er)-.165 E(handling man)100.8 600.6 Q 2.75(yc)-.165 G
(oncurrent connections, and the recei)-2.75 E -.165(ve)-.275 G 2.75(rh)
.165 G(as more memory and CPU c)-2.75 E(ycles)-.165 E -.22(av)100.8
613.6 S(ailable for computation.)-.055 E(In addition, a recei)5.5 E
-.165(ve)-.275 G -.22(r-).165 G(based mechanism is more suitable as a)
.22 E -.22(bu)100.8 626.6 S
(ilding block for multicast congestion control.).22 E(Ho)5.5 E(we)-.275
E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(ti).44 G 2.75(sa)-2.75 G
(lso possible to implement)-2.75 E(TFRC in sender)100.8 639.6 Q
(-based v)-.22 E(ariants, as allo)-.275 E(wed in DCCP')-.275 E 2.75(sC)
-.605 G(ongestion Control ID 3 \(CCID 3\))-2.75 E([RFC4342].)100.8 652.6
Q(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
203.343(idmer Section)-.44 F 2.75(1. [P)2.75 F(age 8])-.165 E 0 Cg EP
%%Page: 9 9
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E/F1 11/Times-Bold@0 SF(2.)72 85
Q/F2 14/Times-Bold@0 SF(Con)5.5 E -.14(ve)-.56 G(ntions).14 E F0(The k)
100.8 101.6 Q .33 -.165(ey w)-.11 H(ords "MUST", "MUST NO).055 E
(T", "REQ)-.44 E(UIRED", "SHALL", "SHALL NO)-.11 E(T",)-.44 E
("SHOULD", "SHOULD NO)100.8 114.6 Q(T", "RECOMMENDED", "MA)-.44 E
(Y", and "OPTION)-1.155 E(AL" in this)-.385 E
(document are to be interpreted as described in [RFC2119].)100.8 127.6 Q
(Appendix A gi)100.8 144.2 Q -.165(ve)-.275 G 2.75(sal).165 G
(ist of technical terms used in this document.)-2.75 E F1(3.)72 170.2 Q
F2(Pr)5.5 E(otocol Mechanism)-.252 E F0 -.165(Fo)100.8 186.8 S 2.75(ri)
.165 G(ts congestion control mechanism, TFRC directly uses a throughput\
 equation for the)-2.75 E(allo)100.8 199.8 Q
(wed sending rate as a function of the loss e)-.275 E -.165(ve)-.275 G
(nt rate and round-trip time.).165 E(In order to)5.5 E(compete f)100.8
212.8 Q(airly with TCP)-.11 E 2.75(,T)-1.221 G
(FRC uses the TCP throughput equation, which roughly)-2.75 E
(describes TCP')100.8 225.8 Q 2.75(ss)-.605 G
(ending rate as a function of the loss e)-2.75 E -.165(ve)-.275 G
(nt rate, round-trip time, and).165 E(se)100.8 238.8 Q(gment size.)-.165
E 1.76 -.88(We d)5.5 H(e\214ne a loss e).88 E -.165(ve)-.275 G
(nt as one or more lost or mark).165 E(ed pack)-.11 E(ets from a windo)
-.11 E(w)-.275 E(of data, where a mark)100.8 251.8 Q(ed pack)-.11 E
(et refers to a congestion indication from Explicit Congestion)-.11 E
(Noti\214cation \(ECN\) [RFC3168].)100.8 264.8 Q
(Generally speaking, TFRC')100.8 281.4 Q 2.75(sc)-.605 G
(ongestion control mechanism w)-2.75 E(orks as follo)-.11 E(ws:)-.275 E
24.95<8354>100.8 298 S(he recei)-24.95 E -.165(ve)-.275 G 2.75(rm).165 G
(easures the loss e)-2.75 E -.165(ve)-.275 G
(nt rate and feeds this information back to the).165 E(sender)129.6 311
Q(.)-.605 E 24.95<8354>100.8 327.6 S(he sender also uses these feedback\
 messages to measure the round-trip time \(R)-24.95 E(TT\).)-.66 E 24.95
<8354>100.8 344.2 S(he loss e)-24.95 E -.165(ve)-.275 G(nt rate and R)
.165 E(TT are then fed into TFRC')-.66 E 2.75(st)-.605 G
(hroughput equation, and the)-2.75 E
(resulting sending rate is limited to at most twice the recei)129.6
357.2 Q .33 -.165(ve r)-.275 H(ate to gi).165 E .33 -.165(ve t)-.275 H
(he allo).165 E(wed)-.275 E(transmit rate X.)129.6 370.2 Q 24.95<8354>
100.8 386.8 S
(he sender then adjusts its transmit rate to match the allo)-24.95 E
(wed transmit rate X.)-.275 E(The dynamics of TFRC are sensiti)100.8
403.4 Q .33 -.165(ve t)-.275 H 2.75(oh).165 G .55 -.275(ow t)-2.75 H
(he measurements are performed and applied.).275 E 1.76 -.88(We r)100.8
416.4 T(ecommend speci\214c mechanisms belo).88 E 2.75(wt)-.275 G 2.75
(op)-2.75 G(erform and apply these measurements.)-2.75 E
(Other mechanisms are possible, b)100.8 429.4 Q
(ut it is important to understand ho)-.22 E 2.75(wt)-.275 G
(he interactions)-2.75 E(between mechanisms af)100.8 442.4 Q
(fect the dynamics of TFRC.)-.275 E F1(3.1.)72 468.4 Q/F3 13
/Times-Bold@0 SF(TCP Thr)5.5 E(oughput Equation)-.234 E F0(An)100.8 485
Q 2.75(yr)-.165 G(ealistic equation gi)-2.75 E
(ving TCP throughput as a function of loss e)-.275 E -.165(ve)-.275 G
(nt rate and R).165 E(TT)-.66 E(should be suitable for use in TFRC.)
100.8 498 Q(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, w).165 H
2.75(en).44 G(ote that the TCP throughput equation)-2.75 E
(used must re\215ect TCP')100.8 511 Q 2.75(sr)-.605 G
(etransmit timeout beha)-2.75 E(vior)-.22 E 2.75(,a)-.44 G 2.75(st)-2.75
G(his dominates TCP throughput at)-2.75 E(higher loss rates.)100.8 524 Q
1.76 -.88(We a)5.5 H
(lso note that the assumptions implicit in the throughput equation).88 E
(about the loss e)100.8 537 Q -.165(ve)-.275 G(nt rate parameter ha).165
E .33 -.165(ve t)-.22 H 2.75(ob).165 G 2.75(ear)-2.75 G
(easonable match to ho)-2.75 E 2.75(wt)-.275 G(he loss rate or)-2.75 E
(loss e)100.8 550 Q -.165(ve)-.275 G(nt rate is actually measured.).165
E(While this match is not perfect for the throughput)5.5 E
(equation and loss rate measurement mechanisms gi)100.8 563 Q -.165(ve)
-.275 G 2.75(nb).165 G(elo)-2.75 E 1.43 -.715(w, i)-.275 H 2.75(np).715
G(ractice the assumptions)-2.75 E(turn out to be close enough.)100.8 576
Q(The throughput equation we currently recommend for TFRC is a slightly\
 simpli\214ed v)100.8 592.6 Q(ersion)-.165 E
(of the throughput equation for Reno TCP from [PFTK98]. Ideally we')
100.8 605.6 Q 2.75(dp)-.55 G(refer a throughput)-2.75 E
(equation based on SA)100.8 618.6 Q(CK TCP)-.44 E 2.75(,b)-1.221 G
(ut no one has yet deri)-2.97 E -.165(ve)-.275 G 2.75(dt).165 G
(he throughput equation for)-2.75 E(SA)100.8 631.6 Q(CK TCP)-.44 E 2.75
(,a)-1.221 G(nd from both simulations and e)-2.75 E(xperiments, the dif)
-.165 E(ferences between the tw)-.275 E(o)-.11 E(equations are relati)
100.8 644.6 Q -.165(ve)-.275 G(ly minor).165 E(.)-.605 E
(The throughput equation is:)100.8 661.2 Q(Handle)72 733 Q(y/Flo)-.165 E
(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 195.093(idmer Section)-.44 F 2.75
(3.1. [P)2.75 F(age 9])-.165 E 0 Cg EP
%%Page: 10 10
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(s)180.55 85 Q
(X_Bps = ----------------------------------------------------------)
100.8 98 Q(R*sqrt\(2*b*p/3\) + \(t_R)122.8 111 Q .396 -.198(TO * \()-.66
H(3*sqrt\(3*b*p/8\)*p*\(1+32*p^2\)\)\)).198 E(Where:)100.8 133.6 Q
(X_Bps is the transmit rate in bytes/second.)129.6 150.2 Q
(\(X_Bps is the same as X_calc in RFC)5.5 E(3448.\))129.6 163.2 Q 2.75
(si)129.6 179.8 S 2.75(st)-2.75 G(he se)-2.75 E(gment size in bytes.)
-.165 E 2.75(Ri)129.6 196.4 S 2.75(st)-2.75 G
(he round trip time in seconds.)-2.75 E 2.75(pi)129.6 213 S 2.75(st)
-2.75 G(he loss e)-2.75 E -.165(ve)-.275 G
(nt rate, between 0 and 1.0, of the number of loss e).165 E -.165(ve)
-.275 G(nts as a fraction).165 E(of the number of pack)129.6 226 Q
(ets transmitted.)-.11 E(t_R)129.6 242.6 Q .396 -.198(TO i)-.66 H 2.75
(st).198 G(he TCP retransmission timeout v)-2.75 E(alue in seconds.)
-.275 E 2.75(bi)129.6 259.2 S 2.75(st)-2.75 G(he maximum number of pack)
-2.75 E(ets ackno)-.11 E(wledged by a single TCP)-.275 E(ackno)129.6
272.2 Q(wledgement.)-.275 E 1.76 -.88(We f)100.8 288.8 T
(urther simplify this by setting t_R).88 E .396 -.198(TO = 4)-.66 H 2.75
(*R. A).198 F(more accurate calculation of t_R)2.75 E .396 -.198(TO i)
-.66 H(s).198 E(possible, b)100.8 301.8 Q(ut e)-.22 E
(xperiments with the current setting ha)-.165 E .33 -.165(ve r)-.22 H
(esulted in reasonable f).165 E(airness with)-.11 E -.165(ex)100.8 314.8
S(isting TCP implementations [W00].).165 E(Another possibility w)5.5 E
(ould be to set t_R)-.11 E .396 -.198(TO =)-.66 H(max\(4R, one second\)\
, to match the recommended minimum of one second on the R)100.8 327.8 Q
-.198(TO)-.66 G([RFC2988].)100.8 340.8 Q(Man)100.8 357.4 Q 2.75(yc)-.165
G(urrent TCP connections use delayed ackno)-2.75 E
(wledgements, sending an)-.275 E(ackno)100.8 370.4 Q(wledgement for e)
-.275 E -.165(ve)-.275 G(ry tw).165 E 2.75(od)-.11 G(ata pack)-2.75 E
(ets recei)-.11 E -.165(ve)-.275 G(d, and thus ha).165 E .33 -.165
(ve a s)-.22 H(ending rate).165 E(modeled by b = 2.)100.8 383.4 Q(Ho)5.5
E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, T).165 H(CP is also allo).44 E
(wed to send an ackno)-.275 E(wledgement for e)-.275 E -.165(ve)-.275 G
(ry).165 E(data pack)100.8 396.4 Q(et, and this w)-.11 E
(ould be modeled by b = 1.)-.11 E(Because man)5.5 E 2.75(yT)-.165 G
(CP implementations do)-2.75 E(not use delayed ackno)100.8 409.4 Q
(wledgements, we recommend b = 1.)-.275 E(In future, dif)100.8 426 Q
(ferent TCP equations may be substituted for this equation.)-.275 E
(The requirement is)5.5 E(that the throughput equation be a reasonable \
approximation of the sending rate of TCP for)100.8 439 Q
(conformant TCP congestion control.)100.8 452 Q
(The throughput equation can also be e)100.8 468.6 Q(xpressed as)-.165 E
(X_Bps =)100.8 487.6 Q(X_pps * s ,)5.5 E
(with X_pps, the sending rate in pack)100.8 506.6 Q(ets per second, gi)
-.11 E -.165(ve)-.275 G 2.75(na).165 G(s)-2.75 E(1)194.3 525.6 Q
(X_pps =)100.8 538.6 Q
(--------------------------------------------------------)5.5 E
(R*sqrt\(2*b*p/3\) + \(t_R)122.8 551.6 Q -.198(TO)-.66 G
(*\(3*sqrt\(3*b*p/8\)*p*\(1+32*p^2\)\)\)).198 E(The parameters s \(se)
100.8 574.2 Q(gment size\), p \(loss e)-.165 E -.165(ve)-.275 G
(nt rate\) and R \(R).165 E(TT\) need to be measured or)-.66 E
(calculated by a TFRC implementation.)100.8 587.2 Q
(The measurement of s is speci\214ed in Section 4.1,)5.5 E(measurement \
of R is speci\214ed in Section 4.3, and measurement of p is speci\214ed\
 in Section)100.8 600.2 Q(5. In the rest of this document data rates ar\
e measured in bytes/second unless otherwise)100.8 613.2 Q(speci\214ed.)
100.8 626.2 Q/F1 11/Times-Bold@0 SF(3.2.)72 652.2 Q/F2 13/Times-Bold@0
SF -.13(Pa)5.5 G(ck).13 E(et Contents)-.13 E F0
(Before specifying the sender and recei)100.8 668.8 Q -.165(ve)-.275 G
2.75(rf).165 G(unctionality)-2.75 E 2.75(,w)-.715 G 2.75(ed)-2.75 G
(escribe the contents of the)-2.75 E(data pack)100.8 681.8 Q
(ets sent by the sender and feedback pack)-.11 E(ets sent by the recei)
-.11 E -.165(ve)-.275 G 3.96 -.605(r. A).165 H 2.75(sT).605 G(FRC will)
-2.75 E(be used along with a transport protocol, we do not specify pack)
100.8 694.8 Q(et formats, as these depend)-.11 E(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44
F 2.75(3.2. [P)2.75 F(age 10])-.165 E 0 Cg EP
%%Page: 11 11
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(on the details of the transport protocol used.)100.8 85 Q/F1 11
/Times-Bold@0 SF 2.75(3.2.1. Data)72 124 R -.11(Pa)2.75 G(ck).11 E(ets)
-.11 E F0(Each data pack)100.8 140.6 Q
(et sent by the data sender contains the follo)-.11 E(wing information:)
-.275 E 24.95<8341>100.8 157.2 S(sequence number)-22.2 E 2.75(.T)-.605 G
(his number is incremented by one for each data pack)-2.75 E(et)-.11 E
2.75(transmitted. The)129.6 170.2 R(\214eld must be suf)2.75 E
(\214ciently lar)-.275 E(ge that it does not wrap causing tw)-.198 E(o)
-.11 E(dif)129.6 183.2 Q(ferent pack)-.275 E
(ets with the same sequence number to be in the recei)-.11 E -.165(ve)
-.275 G(r').165 E 2.75(sr)-.605 G(ecent)-2.75 E(pack)129.6 196.2 Q
(et history at the same time.)-.11 E 24.95<8341>100.8 212.8 S
(timestamp indicating when the pack)-22.2 E(et is sent. W)-.11 E 2.75
(ed)-.88 G(enote by ts_i the timestamp of)-2.75 E(the pack)129.6 225.8 Q
(et with sequence number i.)-.11 E
(The resolution of the timestamp should)5.5 E
(typically be measured in milliseconds.)129.6 238.8 Q
(This timestamp is used by the recei)129.6 255.4 Q -.165(ve)-.275 G 2.75
(rt).165 G 2.75(od)-2.75 G(etermine which losses belong to the same)
-2.75 E(loss e)129.6 268.4 Q -.165(ve)-.275 G 2.75(nt. The).165 F
(timestamp is also echoed by the recei)2.75 E -.165(ve)-.275 G 2.75(rt)
.165 G 2.75(oe)-2.75 G(nable the sender to)-2.75 E
(estimate the round-trip time, for senders that do not sa)129.6 281.4 Q
.33 -.165(ve t)-.22 H(imestamps of transmitted).165 E(data pack)129.6
294.4 Q(ets.)-.11 E 1.76 -.88(We n)129.6 311 T(ote that as an alternati)
.88 E .33 -.165(ve t)-.275 H 2.75(oat).165 G
(imestamp incremented in milliseconds, a)-2.75 E
("timestamp" that increments e)129.6 324 Q -.165(ve)-.275 G
(ry quarter of a round-trip time w).165 E(ould be suf)-.11 E(\214cient)
-.275 E(for determining when losses belong to the same loss e)129.6 337
Q -.165(ve)-.275 G(nt, in the conte).165 E(xt of a)-.165 E
(protocol where this is understood by both sender and recei)129.6 350 Q
-.165(ve)-.275 G .88 -.44(r, a).165 H(nd where the sender).44 E(sa)129.6
363 Q -.165(ve)-.22 G 2.75(st).165 G
(he timestamps of transmitted data pack)-2.75 E(ets.)-.11 E 24.95<8354>
100.8 379.6 S(he sender')-24.95 E 2.75(sc)-.605 G
(urrent estimate of the round trip time. The estimate reported in pack)
-2.75 E(et)-.11 E 2.75(ii)129.6 392.6 S 2.75(sd)-2.75 G(enoted by R_i.)
-2.75 E(The round-trip time estimate is used by the recei)5.5 E -.165
(ve)-.275 G .88 -.44(r, a).165 H(long with).44 E(the timestamp, to dete\
rmine when multiple losses belong to the same loss e)129.6 405.6 Q -.165
(ve)-.275 G(nt.).165 E
(The round-trip time estimate is also used by the recei)129.6 418.6 Q
-.165(ve)-.275 G 2.75(rt).165 G 2.75(od)-2.75 G(etermine the interv)
-2.75 E(al to)-.275 E(use for calculating the recei)129.6 431.6 Q .33
-.165(ve r)-.275 H(ate, and to determine when to send feedback pack).165
E(ets.)-.11 E
(If the sender sends a coarse-grained "timestamp" that increments e)
129.6 448.2 Q -.165(ve)-.275 G(ry quarter of a).165 E
(round-trip time, as discussed abo)129.6 461.2 Q -.165(ve)-.165 G 2.75
(,t).165 G(hen the sender does not need to send its current)-2.75 E
(estimate of the round trip time.)129.6 474.2 Q F1 2.75(3.2.2. F)72
503.8 R(eedback P)-.275 E(ack)-.11 E(ets)-.11 E F0(Each feedback pack)
100.8 520.4 Q(et sent by the data recei)-.11 E -.165(ve)-.275 G 2.75(rc)
.165 G(ontains the follo)-2.75 E(wing information:)-.275 E 24.95<8354>
100.8 537 S(he timestamp of the last data pack)-24.95 E(et recei)-.11 E
-.165(ve)-.275 G(d. W).165 E 2.75(ed)-.88 G(enote this by t_recvdata.)
-2.75 E(If the)5.5 E(last pack)129.6 550 Q(et recei)-.11 E -.165(ve)
-.275 G 2.75(da).165 G 2.75(tt)-2.75 G(he recei)-2.75 E -.165(ve)-.275 G
2.75(rh).165 G(as sequence number i, then t_recvdata = ts_i.)-2.75 E(Th\
is timestamp is used by the sender to estimate the round-trip time, and\
 is only)129.6 563 Q(needed if the sender does not sa)129.6 576 Q .33
-.165(ve t)-.22 H(imestamps of transmitted data pack).165 E(ets.)-.11 E
24.95<8354>100.8 592.6 S
(he amount of time elapsed between the receipt of the last data pack)
-24.95 E(et at the)-.11 E(recei)129.6 605.6 Q -.165(ve)-.275 G .88 -.44
(r, a).165 H(nd the generation of this feedback report. W).44 E 2.75(ed)
-.88 G(enote this by t_delay)-2.75 E(.)-.715 E 24.95<8354>100.8 622.2 S
(he rate at which the recei)-24.95 E -.165(ve)-.275 G 2.75(re).165 G
(stimates that data w)-2.75 E(as recei)-.11 E -.165(ve)-.275 G 2.75(di)
.165 G 2.75(nt)-2.75 G(he pre)-2.75 E(vious round-)-.275 E(trip time.)
129.6 635.2 Q 1.76 -.88(We d)5.5 H(enote this by X_recv).88 E(.)-.715 E
24.95<8354>100.8 651.8 S(he recei)-24.95 E -.165(ve)-.275 G(r').165 E
2.75(sc)-.605 G(urrent estimate of the loss e)-2.75 E -.165(ve)-.275 G
(nt rate p.).165 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E
(ye/W)-.055 E 181.343(idmer Section)-.44 F 2.75(3.2.2. [P)2.75 F
(age 11])-.165 E 0 Cg EP
%%Page: 12 12
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E/F1 11/Times-Bold@0 SF(4.)72 85
Q/F2 14/Times-Bold@0 SF(Data Sender Pr)5.5 E(otocol)-.252 E F0
(The data sender sends a stream of data pack)100.8 101.6 Q
(ets to the data recei)-.11 E -.165(ve)-.275 G 2.75(ra).165 G 2.75(tac)
-2.75 G(ontrolled rate.)-2.75 E(When a feedback pack)100.8 114.6 Q
(et is recei)-.11 E -.165(ve)-.275 G 2.75(df).165 G(rom the data recei)
-2.75 E -.165(ve)-.275 G .88 -.44(r, t).165 H
(he data sender changes its).44 E(sending rate, based on the informatio\
n contained in the feedback report. If the sender does)100.8 127.6 Q
(not recei)100.8 140.6 Q .33 -.165(ve a f)-.275 H(eedback report for fo\
ur round trip times, it cuts its sending rate in half.).165 E(This)5.5 E
(is achie)100.8 153.6 Q -.165(ve)-.275 G 2.75(db).165 G 2.75(ym)-2.75 G
(eans of a timer called the nofeedback timer)-2.75 E(.)-.605 E 1.76 -.88
(We s)100.8 170.2 T(pecify the sender).88 E(-side protocol in the follo)
-.22 E(wing steps:)-.275 E 24.95<834d>100.8 186.8 S
(easurement of the mean se)-24.95 E(gment size being sent.)-.165 E 24.95
<8353>100.8 203.4 S(ender initialization.)-24.95 E 24.95<8354>100.8 220
S(he sender beha)-24.95 E(vior when a feedback pack)-.22 E(et is recei)
-.11 E -.165(ve)-.275 G(d.).165 E 24.95<8354>100.8 236.6 S
(he sender beha)-24.95 E(vior when the nofeedback timer e)-.22 E
(xpires.)-.165 E 24.95<834f>100.8 253.2 S(scillation pre)-24.95 E -.165
(ve)-.275 G(ntion \(optional\)).165 E 24.95<8353>100.8 269.8 S
(cheduling of transmission on non-realtime operating systems.)-24.95 E
F1(4.1.)72 299.4 Q/F3 13/Times-Bold@0 SF(Measuring the Segment Size)5.5
E F0(The parameter s \(se)100.8 316 Q(gment size\) is normally kno)-.165
E(wn to an application.)-.275 E(This may not be so in)5.5 E(tw)100.8 329
Q 2.75(oc)-.11 G(ases:)-2.75 E 24.95<8328>100.8 345.6 S(1\) The se)
-24.95 E(gment size naturally v)-.165 E(aries depending on the data.)
-.275 E(In this case, although)5.5 E(the se)129.6 358.6 Q(gment size v)
-.165 E(aries, that v)-.275 E
(ariation is not coupled to the transmit rate.)-.275 E(The TFRC)5.5 E
(sender can either compute the a)129.6 371.6 Q -.165(ve)-.22 G(rage se)
.165 E(gment size or use the maximum se)-.165 E(gment)-.165 E
(size for the se)129.6 384.6 Q(gment size s.)-.165 E 24.95<8328>100.8
401.2 S(2\) The application needs to change the se)-24.95 E
(gment size rather than the number of)-.165 E(se)129.6 414.2 Q
(gments per second to perform congestion control.)-.165 E(This w)5.5 E
(ould normally be the)-.11 E(case with pack)129.6 427.2 Q
(et audio applications where a \214x)-.11 E(ed interv)-.165 E
(al of time needs to be)-.275 E(represented by each pack)129.6 440.2 Q
2.75(et. Such)-.11 F(applications need to ha)2.75 E .33 -.165(ve a c)
-.22 H(ompletely dif).165 E(ferent)-.275 E -.11(wa)129.6 453.2 S 2.75
(yo).11 G 2.75(fm)-2.75 G(easuring parameters.)-2.75 E -.165(Fo)100.8
469.8 S 2.75(rt).165 G(he \214rst class of applications where the se)
-2.75 E(gment size v)-.165 E(aries depending on the data, the)-.275 E
(sender MA)100.8 482.8 Q 2.75(Ye)-1.155 G(stimate the se)-2.75 E
(gment size s as the a)-.165 E -.165(ve)-.22 G(rage se).165 E
(gment size o)-.165 E -.165(ve)-.165 G 2.75(rt).165 G(he last four loss)
-2.75 E(interv)100.8 495.8 Q 2.75(als. The)-.275 F(sender MA)2.75 E 2.75
(Ya)-1.155 G(lso estimate the a)-2.75 E -.165(ve)-.22 G(rage se).165 E
(gment size o)-.165 E -.165(ve)-.165 G 2.75(rl).165 G(onger time)-2.75 E
(interv)100.8 508.8 Q(als, if so desired.)-.275 E
(The TFRC sender uses the se)5.5 E(gment size s in the throughput)-.165
E(equation, in the setting of the maximum recei)100.8 521.8 Q .33 -.165
(ve r)-.275 H(ate and the minimum and initial sending).165 E
(rates, and in the setting of the nofeedback timer)100.8 534.8 Q(.)-.605
E(The TFRC recei)100.8 551.4 Q -.165(ve)-.275 G 2.75(rm).165 G
(ay use the a)-2.75 E -.165(ve)-.22 G(rage se).165 E
(gment size s in initializing the loss history after)-.165 E
(the \214rst loss e)100.8 564.4 Q -.165(ve)-.275 G(nt, b).165 E
(ut Section 6.3.1 also gi)-.22 E -.165(ve)-.275 G 2.75(sa).165 G 2.75
(na)-2.75 G(lternate procedure that does not use the)-2.75 E -2.475 -.22
(av e)100.8 577.4 T(rage se).22 E(gment size s.)-.165 E(The second clas\
s of applications are discussed separately in a separate document on TF\
RC-)100.8 594 Q(SP)100.8 607 Q 5.5(.F)-1.221 G(or the remainder of this\
 section we assume the sender can estimate the se)-5.665 E(gment size,)
-.165 E(and that congestion control is performed by adjusting the numbe\
r of pack)100.8 620 Q(ets sent per)-.11 E(second.)100.8 633 Q(Handle)72
733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593
(idmer Section)-.44 F 2.75(4.1. [P)2.75 F(age 12])-.165 E 0 Cg EP
%%Page: 13 13
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E/F1 11/Times-Bold@0 SF(4.2.)72
85 Q/F2 13/Times-Bold@0 SF(Sender Initialization)5.5 E F0(The initial v)
100.8 101.6 Q(alues for X \(the allo)-.275 E
(wed sending rate in bytes per second\) and tld \(the T)-.275 E(ime)
-.385 E(Last Doubled during slo)100.8 114.6 Q
(w-start\) are unde\214ned until the)-.275 E 2.75(ya)-.165 G
(re set as described belo)-2.75 E 4.18 -.715(w. I)-.275 H 2.75(ft).715 G
(he)-2.75 E(sender is ready to send data when it does not yet ha)100.8
127.6 Q .33 -.165(ve a r)-.22 H(ound trip sample, the v).165 E
(alue of X is)-.275 E(set to s bytes per second, for se)100.8 140.6 Q
(gment size s, the nofeedback timer is set to e)-.165 E(xpire after tw)
-.165 E(o)-.11 E(seconds, and tld is set either to 0 or to -1.)100.8
153.6 Q(Upon recei)5.5 E(ving a round trip time measurement)-.275 E
(\(e.g., after the \214rst feedback pack)100.8 166.6 Q(et or the SYN e)
-.11 E(xchange, or from a pre)-.165 E(vious connection)-.275 E
([RFC2140]\), tld is set to the current time, and the allo)100.8 179.6 Q
(wed transmit rate X is set to the)-.275 E
(initial_rate, specifed as W_init/R, for W_init based on [RFC3390]:)
100.8 192.6 Q(W_init = min\(4*s, max\(2*s, 4380\)\).)136.8 211.6 Q -.165
(Fo)100.8 230.6 S 2.75(rr).165 G(esponding to the initial feedback pack)
-2.75 E(et, this replaces step \(4\) of Section 4.3 belo)-.11 E -.715
(w.)-.275 G(Appendix B e)100.8 247.2 Q(xplains wh)-.165 E 2.75(yt)-.055
G(he initial v)-2.75 E(alue of TFRC')-.275 E 2.75(sn)-.605 G
(ofeedback timer is set to tw)-2.75 E(o)-.11 E
(seconds, instead of the recommended initial v)100.8 260.2 Q
(alue of three seconds for TCP')-.275 E 2.75(sr)-.605 G(etransmit)-2.75
E(timer from [RFC2988].)100.8 273.2 Q F1(4.3.)72 299.2 Q F2(Sender Beha)
5.5 E(vior When a F)-.325 E(eedback P)-.325 E(ack)-.13 E(et is Recei)
-.13 E -.13(ve)-.13 G(d).13 E F0(The sender kno)100.8 315.8 Q
(ws its current allo)-.275 E
(wed sending rate X, and maintains an estimate of the)-.275 E
(current round trip time R.)100.8 328.8 Q
(The sender also maintains X_recv_set as a small set of recent)5.5 E
(X_recv v)100.8 341.8 Q 2.75(alues. X_recv_set)-.275 F
(is \214rst initialized to contain the v)2.75 E
(alue In\214nity \(or a suitably)-.275 E(lar)100.8 354.8 Q(ge number\).)
-.198 E(The v)5.5 E(ariable recv_limit is de\214ned as the limit on the\
 sending rate that is)-.275 E(computed from the recei)100.8 367.8 Q .33
-.165(ve r)-.275 H 2.75(ate. In).165 F
(this document, in step \(4\) belo)2.75 E 1.43 -.715(w, r)-.275 H
(ecv_limit is speci\214ed).715 E(as twice the maximum v)100.8 380.8 Q
(alue in X_recv_set.)-.275 E(Future documents might specify alternate)
5.5 E -.275(va)100.8 393.8 S(lues for recv_limit.).275 E
(When a feedback pack)100.8 410.4 Q(et is recei)-.11 E -.165(ve)-.275 G
2.75(db).165 G 2.75(yt)-2.75 G(he sender at time t_no)-2.75 E 1.43 -.715
(w, t)-.275 H(he follo).715 E(wing actions)-.275 E(should be performed.)
100.8 423.4 Q 16.887(1\) Calculate)100.8 443.6 R 2.75(an)2.75 G .55
-.275(ew r)-2.75 H(ound trip sample.).275 E(R_sample = \(t_no)129.6
456.6 Q 2.75(w-t)-.275 G(_recvdata\) - t_delay)-2.75 E(.)-.715 E 16.887
(2\) Update)100.8 473.2 R(the round trip time estimate:)2.75 E
(If no feedback has been recei)165.6 492.2 Q -.165(ve)-.275 G 2.75(db)
.165 G(efore)-2.75 E 2.75(R=R)176.6 505.2 S(_sample;)-2.75 E(Else)165.6
518.2 Q 2.75(R=q)176.6 531.2 S(*R + \(1-q\)*R_sample;)-2.75 E
(TFRC is not sensiti)129.6 550.2 Q .33 -.165(ve t)-.275 H 2.75(ot).165 G
(he precise v)-2.75 E(alue for the \214lter constant q, b)-.275 E(ut we)
-.22 E(recommend a def)129.6 563.2 Q(ault v)-.11 E(alue of 0.9.)-.275 E
16.887(3\) Update)100.8 579.8 R(the timeout interv)2.75 E(al:)-.275 E
-1.628 -.66(RT O)165.6 598.8 T 2.75(=m)3.41 G(ax\(4*R, 2*s/X\))-2.75 E
16.887(4\) Update)100.8 621.4 R(the allo)2.75 E
(wed sending rate as follo)-.275 E(ws:)-.275 E(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44
F 2.75(4.3. [P)2.75 F(age 13])-.165 E 0 Cg EP
%%Page: 14 14
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(If \(the entire interv)165.6 85
Q(al co)-.275 E -.165(ve)-.165 G(red by the feedback pack).165 E(et)-.11
E -.11(wa)182.1 98 S 2.75(sad).11 G(ata-limited interv)-2.75 E(al\) {)
-.275 E(Replace X_recv_set contents by In\214nity;)176.6 111 Q 2.75(}E)
165.6 124 S 63.25(lse //)-2.75 F(typical beha)2.75 E(vior)-.22 E
(Update X_recv_set;)176.6 137 Q(recv_limit = 2 * max \(X_recv_set\);)
165.6 150 Q(If \(p > 0\))165.6 163 Q(// congestion a)33 E -.22(vo)-.22 G
(idance phase).22 E(Calculate X_Bps using the TCP throughput equation.)
176.6 176 Q 2.75(X=m)176.6 189 S
(ax\(min\(X_Bps, recv_limit\), s/t_mbi\);)-2.75 E(Else if \(t_no)165.6
202 Q 2.75(w-t)-.275 G(ld >= R\))-2.75 E(// initial slo)11 E(w-start)
-.275 E 2.75(X=m)176.6 215 S
(ax\(min\(2*X, recv_limit\), initial_rate\);)-2.75 E(tld = t_no)176.6
228 Q(w;)-.275 E 16.887(5\) If)100.8 250.6 R(oscillation reduction is u\
sed, calculate the instantaneous transmit rate X_inst,)2.75 E(follo)
129.6 263.6 Q(wing Section 4.5.)-.275 E 16.887(6\) Reset)100.8 280.2 R
(the nofeedback timer to e)2.75 E(xpire after R)-.165 E .396 -.198(TO s)
-.66 H(econds.).198 E(The subroutine for updating X_recv_set belo)100.8
296.8 Q 2.75(wk)-.275 G(eeps a set of X_recv v)-2.86 E(alues recei)-.275
E -.165(ve)-.275 G 2.75(df).165 G(or)-2.75 E(non-data-limited periods o)
100.8 309.8 Q -.165(ve)-.165 G 2.75(rt).165 G(he most recent tw)-2.75 E
2.75(or)-.11 G(ound-trip times.)-2.75 E(Update X_recv_set:)136.8 328.8 Q
(Add X_recv to X_recv_set;)147.8 341.8 Q(Delete from X_recv_set v)147.8
354.8 Q(alues older than)-.275 E(tw)158.8 367.8 Q 2.75(or)-.11 G
(ound-trip times.)-2.75 E 1.76 -.88(We d)100.8 390.4 T
(e\214ne a sender a data-limited an).88 E 2.75(yt)-.165 G
(ime it is not sending as much as it is allo)-2.75 E(wed to send)-.275 E
(\(including unused send credits discussed in Section 4.6\).)100.8 403.4
Q 1.76 -.88(We d)5.5 H(e\214ne an interv).88 E(al as a `data-)-.275 E
(limited interv)100.8 416.4 Q(al' if the sender w)-.275 E
(as data-limited o)-.11 E -.165(ve)-.165 G 2.75(rt).165 G
(he *entire* interv)-2.75 E 2.75(al. The)-.275 F(\214rst `)2.75 E -2.145
(`if ')-.814 F(')-.814 E(condition in step \(4\) pre)100.8 429.4 Q -.165
(ve)-.275 G(nts a sender from ha).165 E
(ving to reduce the sending rate as a result of)-.22 E 2.75(af)100.8
442.4 S(eedback pack)-2.75 E(et reporting the recei)-.11 E .33 -.165
(ve r)-.275 H(ate from a data-limited period.).165 E
(Thus, consider a sender that is sending at its full allo)100.8 459 Q
(wed rate, e)-.275 E(xcept that it is sending)-.165 E(pack)100.8 472 Q
(ets in pairs, rather than sending each pack)-.11 E
(et as soon as it can.)-.11 E(Such a sender is)5.5 E
(considered data-limited part of the time, because it is not al)100.8
485 Q -.11(wa)-.11 G(ys sending pack).11 E(ets as soon as)-.11 E
(it can.)100.8 498 Q(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, a)
.165 H .33 -.165(ny i).44 H(nterv).165 E(al that co)-.275 E -.165(ve)
-.165 G(rs the transmission of at least tw).165 E 2.75(od)-.11 G
(ata pack)-2.75 E(ets is not)-.11 E 2.75(ad)100.8 511 S
(ata-limited interv)-2.75 E(al for this sender)-.275 E(.)-.605 E
(Because X_recv_set is initialized with the v)100.8 527.6 Q
(alue In\214nity)-.275 E 2.75(,r)-.715 G
(ecv_limit is set to In\214nity for the)-2.75 E(\214rst tw)100.8 540.6 Q
2.75(or)-.11 G(ound-trip times of the connection.)-2.75 E
(As a result, the sending rate is not limited by)5.5 E(the recei)100.8
553.6 Q .33 -.165(ve r)-.275 H(ate during that period.).165 E(This a)5.5
E -.22(vo)-.22 G(ids the problem of the sending rate being).22 E
(limited by the v)100.8 566.6 Q
(alue of X_recv from the \214rst feedback pack)-.275 E
(et, which reports only one)-.11 E(se)100.8 579.6 Q(gment recei)-.165 E
-.165(ve)-.275 G 2.75(di).165 G 2.75(nt)-2.75 G
(he last round-trip time,)-2.75 E(Ho)100.8 596.2 Q 2.75(wd)-.275 G
(oes the sender determine the period co)-2.75 E -.165(ve)-.165 G
(red by a feedback pack).165 E 2.75(et? In)-.11 F(general, the)2.75 E
(recei)100.8 609.2 Q -.165(ve)-.275 G 2.75(rw).165 G
(ill be sending a feedback pack)-2.75 E
(et once per round-trip time, so typically the sender)-.11 E
(will be able to determine e)100.8 622.2 Q(xactly the period co)-.165 E
-.165(ve)-.165 G(red by the current feedback pack).165 E(et from)-.11 E
(the pre)100.8 635.2 Q(vious feedback pack)-.275 E 2.75(et. Ho)-.11 F
(we)-.275 E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(nc).44 G
(ases when the pre)-2.75 E(vious feedback pack)-.275 E(et w)-.11 E(as)
-.11 E(lost, or when the recei)100.8 648.2 Q -.165(ve)-.275 G 2.75(rs)
.165 G(ends a feedback pack)-2.75 E
(et early because it detected a lost or ECN-)-.11 E(mark)100.8 661.2 Q
(ed pack)-.11 E(et, the sender will ha)-.11 E .33 -.165(ve t)-.22 H 2.75
(oe).165 G(stimate the interv)-2.75 E(al co)-.275 E -.165(ve)-.165 G
(red by the feedback).165 E(pack)100.8 674.2 Q 2.75(et. As)-.11 F
(speci\214ed in Section 6.2, each feedback pack)2.75 E
(et sent by the recei)-.11 E -.165(ve)-.275 G 2.75(rc).165 G -.165(ove)
-2.75 G(rs a).165 E(round-trip time, for the round-trip time estimate R\
_m maintained by the recei)100.8 687.2 Q -.165(ve)-.275 G 2.75(rR).165 G
(_m)-2.75 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)
-.055 E 189.593(idmer Section)-.44 F 2.75(4.3. [P)2.75 F(age 14])-.165 E
0 Cg EP
%%Page: 15 15
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(seconds before the feedback pack)100.8 85 Q(et w)-.11 E(as sent.)-.11 E
(Note that when p=0, the sender has not yet learned of an)100.8 101.6 Q
2.75(yl)-.165 G(oss e)-2.75 E -.165(ve)-.275 G
(nts, and the sender is in).165 E(the initial slo)100.8 114.6 Q
(w-start phase.)-.275 E(In this initial slo)5.5 E
(w-start phase, the sender can approximately)-.275 E(double the sending\
 rate each round-trip time until a loss occurs. The initial_rate term i\
n)100.8 127.6 Q(step \(4\) gi)100.8 140.6 Q -.165(ve)-.275 G 2.75(sam)
.165 G(inimum allo)-2.75 E(wed sending rate during slo)-.275 E
(w-start of the initial allo)-.275 E(wed)-.275 E(sending rate.)100.8
153.6 Q 1.76 -.88(We n)5.5 H
(ote that if the sender is data-limited during slo).88 E
(w-start, or if the)-.275 E
(connection is limited by the path bandwidth, then the sender isn')100.8
166.6 Q 2.75(tn)-.198 G(ecessarily able to)-2.75 E
(double its sending rate each round-trip time; the sender')100.8 179.6 Q
2.75(ss)-.605 G(ending rate is limited to at most)-2.75 E
(twice the recei)100.8 192.6 Q .33 -.165(ve r)-.275 H
(ate, or at most initial_rate, whiche).165 E -.165(ve)-.275 G 2.75(ri)
.165 G 2.75(sl)-2.75 G(ar)-2.75 E(ger)-.198 E(.)-.605 E
(This is similar to TCP')100.8 218.6 Q 2.75(sb)-.605 G(eha)-2.75 E(vior)
-.22 E 2.75(,w)-.44 G
(here the sending rate is limiting by the rate of incoming)-2.75 E
(ackno)100.8 231.6 Q(wledgement pack)-.275 E
(ets, along with the modi\214cation of the windo)-.11 E 2.75(wi)-.275 G
(ncrease algorithm.)-2.75 E(Thus in TCP')100.8 244.6 Q 2.75(sS)-.605 G
(lo)-2.75 E(w-Start, for the most aggressi)-.275 E .33 -.165(ve c)-.275
H(ase of the TCP recei).165 E -.165(ve)-.275 G 2.75(ra).165 G(ckno)-2.75
E(wledging)-.275 E -2.365 -.275(ev e)100.8 257.6 T(ry data pack).275 E
(et, the TCP sender')-.11 E 2.75(ss)-.605 G
(ending rate is limited to at most twice the rate of these)-2.75 E
(incoming ackno)100.8 270.6 Q(wledgment pack)-.275 E(ets.)-.11 E
(The parameter t_mbi is 64 seconds, and represents the maximum inter)
100.8 287.2 Q(-pack)-.22 E(et back)-.11 E(of)-.11 E(f)-.275 E(interv)
100.8 300.2 Q(al in the persistent absence of feedback.)-.275 E
(Thus, when p > 0 the sender sends at least)5.5 E(one pack)100.8 313.2 Q
(et e)-.11 E -.165(ve)-.275 G(ry 64 seconds.).165 E/F1 11/Times-Bold@0
SF(4.4.)72 339.2 Q/F2 13/Times-Bold@0 SF(Expiration of Nofeedback T)5.5
E(imer)-.234 E F0(This section speci\214es the sender')100.8 355.8 Q
2.75(sr)-.605 G(esponse to a nofeedback timer)-2.75 E 5.5(.T)-.605 G
(he nofeedback timer)-5.5 E(could e)100.8 368.8 Q
(xpire because of an idle period, or because of data or feedback pack)
-.165 E(ets dropped in)-.11 E(the netw)100.8 381.8 Q(ork.)-.11 E
(This section uses the v)100.8 398.4 Q(ariable reco)-.275 E -.165(ve)
-.165 G 2.75(r_rate. When).165 F(the TFRC sender reduces the allo)2.75 E
(wed)-.275 E(sending rate in response to a nofeedback timer)100.8 411.4
Q 2.75(,a)-.44 G(nd the sender has been idle e)-2.75 E -.165(ve)-.275 G
2.75(rs).165 G(ince the)-2.75 E(nofeedback timer w)100.8 424.4 Q
(as set, the allo)-.11 E(wed sending rate is not reduced belo)-.275 E
2.75(wt)-.275 G(he reco)-2.75 E -.165(ve)-.165 G(r_rate.).165 E -.165
(Fo)100.8 437.4 S 2.75(rt).165 G(his document, the reco)-2.75 E -.165
(ve)-.165 G(r_rate is set to the initial_rate.).165 E
(Future documents may)5.5 E -.165(ex)100.8 450.4 S
(plore other possible v).165 E(alues for the reco)-.275 E -.165(ve)-.165
G(r_rate.).165 E(If the nofeedback timer e)100.8 467 Q
(xpires, the sender should perform the follo)-.165 E(wing actions:)-.275
E 16.887(1\) Cut)100.8 483.6 R(the allo)2.75 E
(wed sending rate in half.)-.275 E(If the sender has an R)5.5 E
(TT measurement, the)-.66 E(allo)129.6 496.6 Q(wed sending rate is redu\
ced by setting X_recv; the sending rate is limited to at)-.275 E
(most twice X_recv)129.6 509.6 Q 5.5(.M)-.715 G
(odifying X_recv limits the sending rate, b)-5.5 E(ut also allo)-.22 E
(ws the)-.275 E(sender to slo)129.6 522.6 Q
(w-start, doubling its sending rate each R)-.275 E(TT)-.66 E 2.75(,i)
-.814 G 2.75(ff)-2.75 G(eedback messages)-2.75 E
(resume reporting no losses.)129.6 535.6 Q(If the nofeedback timer e)
129.6 552.2 Q(xpires when the sender does not yet ha)-.165 E .33 -.165
(ve a)-.22 H 2.75(nR).165 G(TT sample)-3.41 E(and has not yet recei)
129.6 565.2 Q -.165(ve)-.275 G 2.75(da).165 G .33 -.165(ny f)-2.75 H
(eedback from the recei).165 E -.165(ve)-.275 G .88 -.44(r, o).165 H
2.75(rw).44 G(hen p == 0, then)-2.75 E(X_recv is not halv)129.6 578.2 Q
(ed, and the sending rate is cut in half directly)-.165 E(.)-.715 E
(If the sender has been idle since this nofeedback timer w)129.6 594.8 Q
(as set and X_recv is less)-.11 E(than the reco)129.6 607.8 Q -.165(ve)
-.165 G(r_rate, then X_recv should not be halv).165 E
(ed in response to the timer)-.165 E -.165(ex)129.6 620.8 S 2.75
(piration. This).165 F(ensures that the allo)2.75 E
(wed sending rate is ne)-.275 E -.165(ve)-.275 G 2.75(rr).165 G
(educed to less than)-2.75 E(half the reco)129.6 633.8 Q -.165(ve)-.165
G(r_rate as a result of an idle period.).165 E
(X_recv_set is the set of recent X_recv v)100.8 650.4 Q 2.75(alues. W)
-.275 F 2.75(eu)-.88 G(se the v)-2.75 E
(ariable timer_limit for the limit)-.275 E
(on the sending rate computed from the e)100.8 663.4 Q
(xpiration of the nofeedback timer)-.165 E(.)-.605 E(Handle)72 733 Q
(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593
(idmer Section)-.44 F 2.75(4.4. [P)2.75 F(age 15])-.165 E 0 Cg EP
%%Page: 16 16
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E/F1 11/Courier@0 SF
(X_recv = max \(X_recv_set\);)129.6 85 Q
(If \(sender does not have an RTT sample and has not received)129.6 98 Q
(any feedback from receiver\))195.6 111 Q
(// We don't have X_Bps or recover_rate yet.)156 124 Q 6.6(X=m)156 137 S
(ax\(X/2, s/t_mbi\);)-6.6 E(Else if \(X_recv < recover_rate, and)129.6
150 Q(sender has been idle ever)175.8 163 Q
(since nofeedback timer was set\))175.8 176 Q
(Timer_limit is not updated;)156 189 Q(Else if \(p==0\))129.6 202 Q
(// We don't have X_Bps yet.)156 215 Q 6.6(X=m)156 228 S
(ax\(X/2, s/t_mbi\);)-6.6 E(Else if \(X_Bps > 2*X_recv\)\))129.6 241 Q
(// 2*X_recv was already limiting the sending rate.)156 254 Q
(timer_limit = X_recv;)156 267 Q(Else)129.6 280 Q
(// The sending rate was limited by X_Bps, not by X_recv.)156 293 Q
(timer_limit = X_Bps/2;)156 306 Q(If \(timer_limit < s/t_mbi\))129.6 319
Q(timer_limit = s/t_mbi;)156 332 Q F0(The term s/t_mbi limits the back)
129.6 354.6 Q(of)-.11 E 2.75(ft)-.275 G 2.75(oo)-2.75 G(ne pack)-2.75 E
(et e)-.11 E -.165(ve)-.275 G(ry 64 seconds.).165 E 16.887(2\) If)100.8
371.2 R(timer_limit has been changed, then do the follo)2.75 E(wing:)
-.275 E F1(If \(timer_limit has been updated\))158.4 390.2 Q
(Replace X_recv_set contents with timer_limit/2.)171.6 403.2 Q
(Recalculate X as in step \(4\) of Section 4.3.)171.6 416.2 Q F0 16.887
(3\) Restart)100.8 438.8 R(the nofeedback timer to e)2.75 E
(xpire after max\(4*R, 2*s/X\) seconds.)-.165 E(Note that when the allo)
100.8 455.4 Q
(wed sending rate is limited after an idle period, it is ne)-.275 E
-.165(ve)-.275 G 2.75(rr).165 G(educed)-2.75 E(belo)100.8 468.4 Q 2.75
(wh)-.275 G(alf the reco)-2.75 E -.165(ve)-.165 G(r_rate.).165 E
(If the sender has been data-limited b)100.8 485 Q
(ut not idle since the nofeedback timer w)-.22 E(as set, it is)-.11 E
(possible that the nofeedback timer e)100.8 498 Q
(xpired because data or feedback pack)-.165 E(ets were dropped)-.11 E
(in the netw)100.8 511 Q 2.75(ork. In)-.11 F
(this case, the nofeedback timer is the backup mechanism for the sender)
2.75 E(to detect these losses, similar to the retransmit timer in TCP)
100.8 524 Q(.)-1.221 E
(Note that when the sender stops sending, the recei)100.8 540.6 Q -.165
(ve)-.275 G 2.75(rw).165 G(ill stop sending feedback.)-2.75 E(When)5.5 E
(the sender')100.8 553.6 Q 2.75(sn)-.605 G(ofeedback timer e)-2.75 E
(xpires, the sender could use the procedure abo)-.165 E .33 -.165(ve t)
-.165 H 2.75(ol).165 G(imit)-2.75 E(the sending rate.)100.8 566.6 Q
(If the sender subsequently starts to send ag)5.5 E
(ain, X_recv_set will be used to)-.055 E
(limit the transmit rate, and slo)100.8 579.6 Q(wstart beha)-.275 E
(vior will occur until the transmit rate reaches)-.22 E(X_Bps.)100.8
592.6 Q(The TFRC sender')100.8 609.2 Q 2.75(sr)-.605 G
(eduction of the allo)-2.75 E
(wed sending rate after the nofeedback timer)-.275 E -.165(ex)100.8
622.2 S(pires is similar to TCP').165 E 2.75(sr)-.605 G
(eduction of the congestion windo)-2.75 E 2.75(wc)-.275 G
(wnd after each R)-2.75 E -.198(TO)-.66 G
(seconds of an idle period, for TCP with Congestion W)100.8 635.2 Q
(indo)-.44 E 2.75(wV)-.275 G(alidation [RFC2861].)-3.971 E/F2 11
/Times-Bold@0 SF(4.5.)72 661.2 Q/F3 13/Times-Bold@0 SF
(Reducing Oscillations)5.5 E F0 1.76 -.88(To r)100.8 677.8 T
(educe oscillations in queueing delay and sending rate in en).88 E
(vironments with a lo)-.44 E(w)-.275 E(de)100.8 690.8 Q
(gree of statistical multiple)-.165 E
(xing at the congested link, it can be useful for the sender to)-.165 E
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
189.593(idmer Section)-.44 F 2.75(4.5. [P)2.75 F(age 16])-.165 E 0 Cg EP
%%Page: 17 17
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(reduce the transmit rate as the queuing delay \(and hence R)100.8 85 Q
(TT\) increases.)-.66 E 1.76 -.88(To d)5.5 H 2.75(ot).88 G(his the)-2.75
E(sender maintains R_sqmean, a long-term estimate of the square root of\
 the R)100.8 98 Q(TT)-.66 E 2.75(,a)-.814 G(nd)-2.75 E
(modi\214es its sending rate depending on ho)100.8 111 Q 2.75(wt)-.275 G
(he square root of R_sample, the most recent)-2.75 E(sample of the R)
100.8 124 Q(TT)-.66 E 2.75(,d)-.814 G(if)-2.75 E
(fers from the long-term estimate.)-.275 E
(The long-term estimate R_sqmean)5.5 E(is set as follo)100.8 137 Q(ws:)
-.275 E(If no feedback has been recei)136.8 156 Q -.165(ve)-.275 G 2.75
(db).165 G(efore)-2.75 E(R_sqmean = sqrt\(R_sample\);)147.8 169 Q(Else)
136.8 182 Q(R_sqmean = q2*R_sqmean + \(1-q2\)*sqrt\(R_sample\);)147.8
195 Q(Thus R_sqmean gi)100.8 214 Q -.165(ve)-.275 G 2.75(st).165 G(he e)
-2.75 E(xponentially weighted mo)-.165 E(ving a)-.165 E -.165(ve)-.22 G
(rage of the square root of the).165 E -.66(RT)100.8 227 S 2.75(Ts).66 G
2.75(amples. The)-2.75 F(constant q2 should be set similarly to q, the \
constant used in the round)2.75 E(trip time estimate R.)100.8 240 Q 1.76
-.88(We r)5.5 H(ecommend a v).88 E(alue of 0.9 as the def)-.275 E
(ault for q2.)-.11 E(When sqrt\(R_sample\) is greater than R_sqmean the\
n the current round-trip time is greater)100.8 256.6 Q
(than the long-term a)100.8 269.6 Q -.165(ve)-.22 G
(rage, implying that queueing delay is probably increasing.).165 E
(In this)5.5 E(case, the transmit rate is decreased to minimize oscilla\
tions in queueing delay)100.8 282.6 Q(.)-.715 E
(The sender obtains the base allo)100.8 299.2 Q
(wed transmit rate, X, as described in step \(4\) of Section 4.3)-.275 E
(abo)100.8 312.2 Q -.165(ve)-.165 G 5.5(.I).165 G 2.75(tt)-5.5 G(hen ca\
lculates a modi\214ed instantaneous transmit rate X_inst, as follo)-2.75
E(ws:)-.275 E(X_inst = X * R_sqmean / sqrt\(R_sample\);)136.8 331.2 Q
(If \(p > 0\))136.8 344.2 Q(// congestion a)44 E -.22(vo)-.22 G
(idance phase).22 E(X_inst = max\(X_inst, s/t_mbi\))147.8 357.2 Q
(Else if \(t_no)136.8 370.2 Q 2.75(w-t)-.275 G(ld >= R\))-2.75 E
(// initial slo)22 E(w-start)-.275 E(X_inst = max\(X_inst, s/R\))147.8
383.2 Q(Because we are using square roots, there is generally only a mo\
derate dif)100.8 402.2 Q(ference between)-.275 E
(the instantaneous transmit rate X_inst and the allo)100.8 415.2 Q
(wed transmit rate X.)-.275 E -.165(Fo)5.5 G 2.75(re).165 G
(xample, in a)-2.915 E(some)100.8 428.2 Q(what e)-.275 E
(xtreme case when the current R)-.165 E
(TT sample R_sample is twice as lar)-.66 E(ge as the)-.198 E
(long-term a)100.8 441.2 Q -.165(ve)-.22 G(rage, then sqrt\(R_sample\) \
will be roughly 1.44 times R_sqmean, and the).165 E(allo)100.8 454.2 Q
(wed transmit rate will be reduced by a f)-.275 E(actor of roughly 0.7.)
-.11 E(Note: This modi\214cation for reducing oscillatory beha)100.8
470.8 Q(vior is not al)-.22 E -.11(wa)-.11 G(ys needed, especially).11 E
(if the de)100.8 483.8 Q(gree of statistical multiple)-.165 E
(xing in the netw)-.165 E(ork is high.)-.11 E(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, i).165 H 2.75(tS).44 G(HOULD be)-2.75 E
(implemented because it mak)100.8 496.8 Q(es TFRC beha)-.11 E .33 -.165
(ve b)-.22 H(etter in en).165 E(vironments with a lo)-.44 E 2.75(wl)
-.275 G -2.365 -.275(ev e)-2.75 H 2.75(lo).275 G(f)-2.75 E
(statistical multiple)100.8 509.8 Q 2.75(xing. The)-.165 F
(performance of this modi\214cation is illustrated in Section 3.1.3)2.75
E(of [FHPW00].)100.8 522.8 Q
(If it is not implemented, we recommend using a v)5.5 E(ery lo)-.165 E
2.75(wv)-.275 G(alue of the)-3.025 E(weight q for the a)100.8 535.8 Q
-.165(ve)-.22 G(rage round-trip time.).165 E/F1 11/Times-Bold@0 SF(4.6.)
72 561.8 Q/F2 13/Times-Bold@0 SF(Scheduling of P)5.5 E(ack)-.13 E(et T)
-.13 E(ransmissions)-.962 E F0(As TFRC is rate-based, and as operating \
systems typically cannot schedule e)100.8 578.4 Q -.165(ve)-.275 G(nts)
.165 E(precisely)100.8 591.4 Q 2.75(,i)-.715 G 2.75(ti)-2.75 G 2.75(sn)
-2.75 G(ecessary to be opportunistic about sending data pack)-2.75 E
(ets so that the correct)-.11 E -2.475 -.22(av e)100.8 604.4 T
(rage rate is maintained despite the coarse-grain or irre).22 E
(gular scheduling of the operating)-.165 E 2.75(system. T)100.8 617.4 R
2.75(oh)-.88 G(elp maintain the correct a)-2.75 E -.165(ve)-.22 G
(rage sending rate, the TFRC sender may send).165 E(some pack)100.8
630.4 Q(ets before their nominal send time.)-.11 E
(In addition, the scheduling of pack)100.8 647 Q
(et transmissions controls the allo)-.11 E(wed b)-.275 E(urstiness of)
-.22 E(senders after an idle or data-limited period.)100.8 660 Q(Allo)
5.5 E(wing the TFRC sender to accumulate)-.275 E
(sending `credits' for past unused send times allo)100.8 673 Q
(ws the TFRC sender to send a b)-.275 E(urst or data)-.22 E
(after an idle period.)100.8 686 Q(As a comparison with TCP)5.5 E 2.75
(,T)-1.221 G(CP may send up to a round-trip time')-2.75 E(s)-.605 E
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
189.593(idmer Section)-.44 F 2.75(4.6. [P)2.75 F(age 17])-.165 E 0 Cg EP
%%Page: 18 18
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E -.11(wo)100.8 85 S(rth of pack)
.11 E(ets in a single b)-.11 E(urst, b)-.22 E(ut ne)-.22 E -.165(ve)
-.275 G 2.75(rm).165 G 2.75(ore. As)-2.75 F -.165(ex)2.75 G
(amples, for TCP b).165 E(ursts can be)-.22 E(sent when an A)100.8 98 Q
(CK arri)-.44 E -.165(ve)-.275 G 2.75(sa).165 G(ckno)-2.75 E
(wledging a windo)-.275 E 2.75(wo)-.275 G 2.75(fd)-2.75 G
(ata, or when a data-limited sender)-2.75 E(,)-.44 E
(after a delay of nearly a round-trip time, suddenly has a windo)100.8
111 Q 2.75(wo)-.275 G 2.75(fd)-2.75 G(ata to send.)-2.75 E 1.76 -.88
(To l)100.8 127.6 T(imit b).88 E
(urstiness, a TFRC implementation MUST pre)-.22 E -.165(ve)-.275 G(nt b)
.165 E(ursts of arbitrary size.)-.22 E(This)5.5 E
(limit MUST be less than or equal to one round-trip time')100.8 140.6 Q
2.75(sw)-.605 G(orth of pack)-2.86 E 2.75(ets. A)-.11 F(TFRC)2.75 E
(implementation MA)100.8 153.6 Q 2.75(Yl)-1.155 G(imit b)-2.75 E
(ursts to less than a round-trip time')-.22 E 2.75(sw)-.605 G
(orth of pack)-2.86 E(ets, if so)-.11 E 2.75(desired. Ho)100.8 166.6 R
(we)-.275 E -.165(ve)-.275 G .88 -.44(r, w).165 H 2.75(en).44 G
(ote that such limits also constrain TFRC')-2.75 E 2.75(sp)-.605 G
(erformance be)-2.75 E(yond the)-.165 E(case for the current TCP)100.8
179.6 Q(.)-1.221 E 2.75(At)100.8 196.2 S
(ypical sending loop will calculate the correct inter)-2.75 E(-pack)-.22
E(et interv)-.11 E(al, t_ipi, as follo)-.275 E(ws:)-.275 E
(t_ipi = s/X_inst;)136.8 215.2 Q(Let t_no)100.8 234.2 Q 2.75(wb)-.275 G
2.75(et)-2.75 G(he current time and i be a natural number)-2.75 E 2.75
(,i=0)-.44 G 2.75(,1)-2.75 G 2.75(,.)-2.75 G(.., with t_i the nominal)
-2.75 E(send time for the i-th pack)100.8 247.2 Q 2.75(et. Then)-.11 F
(the nominal send time t_\(i+1\) deri)2.75 E -.165(ve)-.275 G 2.75(sr)
.165 G(ecursi)-2.75 E -.165(ve)-.275 G(ly as).165 E(t_0 = t_no)142.3
266.2 Q -.715(w,)-.275 G(t_\(i+1\) = t_i + t_ipi.)142.3 279.2 Q -.165
(Fo)100.8 298.2 S 2.75(rT).165 G(FRC senders allo)-2.75 E
(wed to accumulate sending `credits' for unused sent time o)-.275 E
-.165(ve)-.165 G 2.75(rt).165 G(he)-2.75 E(last T seconds, the sender w)
100.8 311.2 Q(ould be allo)-.11 E
(wed to use unused nominal sent times t_j for t_j <)-.275 E(no)100.8
324.2 Q 2.75(w-T)-.275 G 5.5(.W)-3.564 G 2.75(er)-6.38 G
(ecommend T set to the round-trip time.)-2.75 E/F1 11/Times-Bold@0 SF
2.75(4.6.1. Sending)72 350.2 R -.11(Pa)2.75 G(ck).11 E(ets Bef)-.11 E
(or)-.275 E 2.75(et)-.198 G(heir Nominal Send T)-2.75 E(ime)-.198 E F0(\
Let t_gran be the scheduling timer granularity of the operating system.)
100.8 366.8 Q(If the operating)5.5 E(system has a coarse timer granular\
ity or otherwise cannot support short t_ipi interv)100.8 379.8 Q
(als, then)-.275 E(either the TFRC sender will be restricted to a sendi\
ng rate of at most 1 pack)100.8 392.8 Q(et e)-.11 E -.165(ve)-.275 G
(ry t_gran).165 E(seconds, or the TFRC sender must be allo)100.8 405.8 Q
(wed to send short b)-.275 E(ursts of pack)-.22 E 2.75(ets. In)-.11 F
(addition)2.75 E(to allo)100.8 418.8 Q(wing the sender to accumulate se\
nding `credits' for past unused send times, it can be)-.275 E
(useful to allo)100.8 431.8 Q 2.75(wt)-.275 G(he sender to send a pack)
-2.75 E(et before its scheduled send time, as described in)-.11 E
(the section belo)100.8 444.8 Q -.715(w.)-.275 G 2.75(Ap)100.8 461.4 S
(arameter t_delta MA)-2.75 E 2.75(Yb)-1.155 G 2.75(eu)-2.75 G
(sed to allo)-2.75 E 2.75(wap)-.275 G(ack)-2.75 E
(et to be sent before its nominal send)-.11 E 2.75(time. Consider)100.8
474.4 R(an application that becomes idle and requests re-scheduling for\
 time t_i =)2.75 E
(t_\(i-1\) + t_ipi, for t_\(i-1\) the send time for the pre)100.8 487.4
Q(vious pack)-.275 E 2.75(et. When)-.11 F(the application is re-)2.75 E
(scheduled, it checks the current time, t_no)100.8 500.4 Q 4.18 -.715
(w. I)-.275 H 2.75(f\().715 G(t_no)-2.75 E 2.75(w>t)-.275 G
(_i - t_delta\) then pack)-2.75 E(et i is sent.)-.11 E
(When the nominal send time, t_i, of the ne)100.8 513.4 Q(xt pack)-.165
E(et is calculated, it may already be the)-.11 E(case that t_no)100.8
526.4 Q 2.75(w>t)-.275 G(_i - t_delta.)-2.75 E(In such a case the pack)
5.5 E(et w)-.11 E(ould be sent immediately)-.11 E(.)-.715 E
(In order to send at most one pack)100.8 543 Q
(et before its nominal send time, and ne)-.11 E -.165(ve)-.275 G 2.75
(rt).165 G 2.75(os)-2.75 G(end a)-2.75 E(pack)100.8 556 Q(et more than \
a round-trip time before its nominal send time the parameter t_delta)
-.11 E -.11(wo)100.8 569 S(uld be set as follo).11 E(ws:)-.275 E
(t_delta = min\(t_ipi, t_gran, rtt\)/2;)136.8 588 Q
(The scheduling granularity t_gran is 10ms on man)100.8 607 Q 2.75(yU)
-.165 G(nix systems.)-2.75 E(If t_gran is not kno)5.5 E(wn,)-.275 E 2.75
(av)100.8 620 S(alue of 10ms could be assumed.)-3.025 E(As an e)100.8
636.6 Q(xample, consider a TFRC \215o)-.165 E 2.75(ww)-.275 G
(ith an allo)-2.75 E(wed sending rate X of 10 pack)-.275 E(ets per)-.11
E(round-trip time, a round-trip time of 100 ms, a system with a schedul\
ing granularity t_gran)100.8 649.6 Q(of 10 ms, and the ability to accum\
ulate unused sending credits for a round-trip time.)100.8 662.6 Q
(In this)5.5 E(case, t_ipi is 1 ms.)100.8 675.6 Q(The TFRC sender w)5.5
E(ould be allo)-.11 E(wed to send pack)-.275 E(ets 0.5 ms before their)
-.11 E(nominal sending time, and w)100.8 688.6 Q(ould be allo)-.11 E
(wed to sa)-.275 E .33 -.165(ve u)-.22 H
(nused sending credits for 100 ms.).165 E(Handle)72 733 Q(y/Flo)-.165 E
(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 181.343(idmer Section)-.44 F 2.75
(4.6.1. [P)2.75 F(age 18])-.165 E 0 Cg EP
%%Page: 19 19
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(The scheduling granularity of 10 ms w)100.8 85 Q
(ould not signi\214cantly af)-.11 E(fect the performance of the)-.275 E
(connection.)100.8 98 Q(As a dif)100.8 114.6 Q(ferent e)-.275 E
(xample, consider a TFRC \215o)-.165 E 2.75(ww)-.275 G
(ith a scheduling granularity less than the)-2.75 E
(round-trip time, for e)100.8 127.6 Q
(xample, with a round-trip time of 0.1 ms and a system with a)-.165 E(s\
cheduling granularity of 1 ms, and with the ability to accumulate unuse\
d sending credits)100.8 140.6 Q(for a round-trip time.)100.8 153.6 Q
(The TFRC sender w)5.5 E(ould be allo)-.11 E(wed to sa)-.275 E .33 -.165
(ve u)-.22 H(nused sending credits).165 E(for 0.1 ms.)100.8 166.6 Q
(If the scheduling granularity *did not* af)5.5 E(fect the sender')-.275
E 2.75(sr)-.605 G(esponse to an)-2.75 E(incoming feedback pack)100.8
179.6 Q(et, then the TFRC sender w)-.11 E(ould be able to send an R)-.11
E(TT of data \(as)-.66 E(determined by the allo)100.8 192.6 Q
(wed sending rate\) each R)-.275 E(TT)-.66 E 2.75(,i)-.814 G 2.75(nr)
-2.75 G(esponse to incoming feedback)-2.75 E(pack)100.8 205.6 Q 2.75
(ets. In)-.11 F(this case, the coarse scheduling granularity w)2.75 E
(ould not signi\214cantly reduce the)-.11 E(sending rate, b)100.8 218.6
Q(ut the sending rate w)-.22 E(ould be b)-.11 E(ursty)-.22 E 2.75(,w)
-.715 G(ith a round-trip time of data sent in)-2.75 E
(response to each feedback pack)100.8 231.6 Q(et.)-.11 E(Ho)100.8 248.2
Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, p).165 H(erformance w).44 E
(ould be dif)-.11 E
(ferent in this case if the operating system scheduling)-.275 E
(granularity af)100.8 261.2 Q(fected the sender')-.275 E 2.75(sr)-.605 G
(esponse to feedback pack)-2.75 E(ets as well as the general)-.11 E
(scheduling of the sender)100.8 274.2 Q 2.75(,I)-.44 G 2.75(nt)-2.75 G
(his case the sender')-2.75 E 2.75(sp)-.605 G(erformance w)-2.75 E
(ould be se)-.11 E -.165(ve)-.275 G(rely limited).165 E(by the scheduli\
ng granularity being less than the round-trip time, with the sender abl\
e to)100.8 287.2 Q(send an R)100.8 300.2 Q(TT of data, at the allo)-.66
E(wed sending rate, at most once e)-.275 E -.165(ve)-.275 G(ry 1 ms.)
.165 E(This restriction)5.5 E(of the sending rate is an una)100.8 313.2
Q -.22(vo)-.22 G(idable consequence of allo).22 E(wing b)-.275 E
(urstiness of at most a)-.22 E(round-trip time of data.)100.8 326.2 Q/F1
11/Times-Bold@0 SF(5.)72 352.2 Q/F2 14/Times-Bold@0 SF
(Calculation of the Loss Ev)5.5 E(ent Rate \(p\))-.14 E F0
(Obtaining an accurate and stable measurement of the loss e)100.8 368.8
Q -.165(ve)-.275 G(nt rate is of primary).165 E
(importance for TFRC. Loss rate measurement is performed at the recei)
100.8 381.8 Q -.165(ve)-.275 G .88 -.44(r, b).165 H(ased on the).44 E
(detection of lost or mark)100.8 394.8 Q(ed pack)-.11 E
(ets from the sequence numbers of arri)-.11 E(ving pack)-.275 E(ets. W)
-.11 E(e)-.88 E
(describe this process before describing the rest of the recei)100.8
407.8 Q -.165(ve)-.275 G 2.75(rp).165 G 2.75(rotocol. If)-2.75 F
(the recei)2.75 E -.165(ve)-.275 G 2.75(rh).165 G(as)-2.75 E
(not yet detected a lost or mark)100.8 420.8 Q(ed pack)-.11 E
(et, then the recei)-.11 E -.165(ve)-.275 G 2.75(rd).165 G(oesn')-2.75 E
2.75(tc)-.198 G(alculate the loss e)-2.75 E -.165(ve)-.275 G(nt).165 E
(rate, b)100.8 433.8 Q(ut reports a loss e)-.22 E -.165(ve)-.275 G
(nt rate of zero.).165 E F1(5.1.)72 459.8 Q/F3 13/Times-Bold@0 SF
(Detection of Lost or Mark)5.5 E(ed P)-.13 E(ack)-.13 E(ets)-.13 E F0
(TFRC assumes that all pack)100.8 476.4 Q
(ets contain a sequence number that is incremented by one for)-.11 E
(each pack)100.8 489.4 Q(et that is sent.)-.11 E -.165(Fo)5.5 G 2.75(rt)
.165 G(he purposes of this speci\214cation, we require that if a lost)
-2.75 E(pack)100.8 502.4 Q
(et is retransmitted, the retransmission is gi)-.11 E -.165(ve)-.275 G
2.75(nan).165 G .55 -.275(ew s)-2.75 H
(equence number that is the latest).275 E(in the transmission sequence,\
 and not the same sequence number as the pack)100.8 515.4 Q(et that w)
-.11 E(as)-.11 E 2.75(lost. If)100.8 528.4 R 2.75(at)2.75 G(ransport pr\
otocol has the requirement that it must retransmit with the original)
-2.75 E(sequence number)100.8 541.4 Q 2.75(,t)-.44 G
(hen the transport protocol designer must \214gure out ho)-2.75 E 2.75
(wt)-.275 G 2.75(od)-2.75 G(istinguish)-2.75 E
(delayed from retransmitted pack)100.8 554.4 Q(ets and ho)-.11 E 2.75
(wt)-.275 G 2.75(od)-2.75 G(etect lost retransmissions.)-2.75 E
(The recei)100.8 571 Q -.165(ve)-.275 G 2.75(rm).165 G
(aintains a data structure that k)-2.75 E(eeps track of which pack)-.11
E(ets ha)-.11 E .33 -.165(ve a)-.22 H(rri).165 E -.165(ve)-.275 G 2.75
(da).165 G(nd)-2.75 E(which are missing.)100.8 584 Q -.165(Fo)5.5 G 2.75
(rt).165 G
(he purposes of speci\214cation, we assume that the data structure)-2.75
E(consists of a list of pack)100.8 597 Q(ets that ha)-.11 E .33 -.165
(ve a)-.22 H(rri).165 E -.165(ve)-.275 G 2.75(da).165 G
(long with the recei)-2.75 E -.165(ve)-.275 G 2.75(rt).165 G
(imestamp when each)-2.75 E(pack)100.8 610 Q(et w)-.11 E(as recei)-.11 E
-.165(ve)-.275 G 2.75(d. In).165 F
(practice this data structure will normally be stored in a more)2.75 E
(compact representation, b)100.8 623 Q
(ut this is implementation-speci\214c.)-.22 E(The loss of a pack)100.8
639.6 Q(et is detected by the arri)-.11 E -.275(va)-.275 G 2.75(lo).275
G 2.75(fa)-2.75 G 2.75(tl)-2.75 G(east NDUP)-2.75 E -.44(AC)-1.012 G
2.75(Kp).44 G(ack)-2.75 E(ets with a higher)-.11 E
(sequence number than the lost pack)100.8 652.6 Q(et, for NDUP)-.11 E
-.44(AC)-1.012 G 2.75(Ks).44 G(et to 3.)-2.75 E(The requirement for)5.5
E(NDUP)100.8 665.6 Q -.44(AC)-1.012 G 2.75(Ks).44 G(ubsequent pack)-2.75
E(ets is the same as with TCP)-.11 E 2.75(,a)-1.221 G(nd is to mak)-2.75
E 2.75(eT)-.11 G(FRC more rob)-2.75 E(ust)-.22 E
(in the presence of reordering.)100.8 678.6 Q(In contrast to TCP)5.5 E
2.75(,i)-1.221 G 2.75(fap)-2.75 G(ack)-2.75 E(et arri)-.11 E -.165(ve)
-.275 G 2.75(sl).165 G(ate \(after NDUP)-2.75 E -.44(AC)-1.012 G(K).44 E
(subsequent pack)100.8 691.6 Q(ets arri)-.11 E -.165(ve)-.275 G
(d\) in TFRC, the late pack).165 E(et can \214ll the hole in TFRC')-.11
E 2.75(sr)-.605 G(eception)-2.75 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)
-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75
(5.1. [P)2.75 F(age 19])-.165 E 0 Cg EP
%%Page: 20 20
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(record, and the recei)100.8 85
Q -.165(ve)-.275 G 2.75(rc).165 G(an recalculate the loss e)-2.75 E
-.165(ve)-.275 G(nt rate.).165 E(Future v)5.5 E(ersions of TFRC might)
-.165 E(mak)100.8 98 Q 2.75(et)-.11 G(he requirement for NDUP)-2.75 E
-.44(AC)-1.012 G 2.75(Ks).44 G(ubsequent pack)-2.75 E(ets adapti)-.11 E
.33 -.165(ve b)-.275 H(ased on e).165 E(xperienced)-.165 E(pack)100.8
111 Q(et reordering, b)-.11 E
(ut we do not specify such a mechanism here.)-.22 E -.165(Fo)100.8 127.6
S 2.75(ra).165 G 2.75(nE)-2.75 G(CN-capable connection, a mark)-2.75 E
(ed pack)-.11 E(et is detected as a congestion e)-.11 E -.165(ve)-.275 G
(nt as soon).165 E(as it arri)100.8 140.6 Q -.165(ve)-.275 G
(s, without ha).165 E(ving to w)-.22 E(ait for the arri)-.11 E -.275(va)
-.275 G 2.75(lo).275 G 2.75(fs)-2.75 G(ubsequent pack)-2.75 E(ets.)-.11
E/F1 11/Times-Bold@0 SF(5.2.)72 179.6 Q/F2 13/Times-Bold@0 SF -.962(Tr)
5.5 G(anslation fr).962 E(om Loss History to Loss Ev)-.234 E(ents)-.13 E
F0(TFRC requires that the loss fraction be rob)100.8 196.2 Q(ust to se)
-.22 E -.165(ve)-.275 G(ral consecuti).165 E .33 -.165(ve p)-.275 H(ack)
.165 E(ets lost or mark)-.11 E(ed)-.11 E(where those pack)100.8 209.2 Q
(ets are part of the same loss e)-.11 E -.165(ve)-.275 G 2.75(nt. This)
.165 F(is similar to TCP)2.75 E 2.75(,w)-1.221 G(hich)-2.75 E
(\(typically\) only performs one halving of the congestion windo)100.8
222.2 Q 2.75(wd)-.275 G(uring an)-2.75 E 2.75(ys)-.165 G(ingle R)-2.75 E
(TT)-.66 E(.)-.814 E(Thus the recei)100.8 235.2 Q -.165(ve)-.275 G 2.75
(rn).165 G(eeds to map the pack)-2.75 E(et loss history into a loss e)
-.11 E -.165(ve)-.275 G(nt record, where a loss).165 E -2.365 -.275
(ev e)100.8 248.2 T(nt is one or more pack).275 E(ets lost or mark)-.11
E(ed in an R)-.11 E(TT)-.66 E 5.5(.T)-.814 G 2.75(op)-6.38 G
(erform this mapping, the)-2.75 E(recei)100.8 261.2 Q -.165(ve)-.275 G
2.75(rn).165 G(eeds to kno)-2.75 E 2.75(wt)-.275 G(he R)-2.75 E
(TT to use, and this is supplied periodically by the sender)-.66 E(,)
-.44 E(typically as control information piggy-back)100.8 274.2 Q
(ed onto a data pack)-.11 E 2.75(et. TFRC)-.11 F(is not sensiti)2.75 E
.33 -.165(ve t)-.275 H(o).165 E(ho)100.8 287.2 Q 2.75(wt)-.275 G(he R)
-2.75 E(TT measurement sent to the recei)-.66 E -.165(ve)-.275 G 2.75
(ri).165 G 2.75(sm)-2.75 G(ade, b)-2.75 E(ut we recommend using the)-.22
E(sender')100.8 300.2 Q 2.75(sc)-.605 G(alculated R)-2.75 E(TT)-.66 E
2.75(,R)-.814 G 2.75(,\()-2.75 G(see Section 4.3\) for this purpose.)
-2.75 E 1.76 -.88(To d)100.8 316.8 T(etermine whether a lost or mark).88
E(ed pack)-.11 E(et should start a ne)-.11 E 2.75(wl)-.275 G(oss e)-2.75
E -.165(ve)-.275 G(nt, or be counted).165 E(as part of an e)100.8 329.8
Q(xisting loss e)-.165 E -.165(ve)-.275 G
(nt, we need to compare the sequence numbers and timestamps).165 E
(of the pack)100.8 342.8 Q(ets that arri)-.11 E -.165(ve)-.275 G 2.75
(da).165 G 2.75(tt)-2.75 G(he recei)-2.75 E -.165(ve)-.275 G 3.96 -.605
(r. F).165 H(or a mark).44 E(ed pack)-.11 E(et S_ne)-.11 E 1.43 -.715
(w, i)-.275 H(ts reception time).715 E(T_ne)100.8 355.8 Q 2.75(wc)-.275
G(an be noted directly)-2.75 E 5.5(.F)-.715 G(or a lost pack)-5.665 E
(et, we can interpolate to infer the nominal)-.11 E("arri)100.8 368.8 Q
-.275(va)-.275 G 2.75(lt).275 G 2.75(ime". Assume:)-2.75 F
(S_loss is the sequence number of a lost pack)129.6 385.4 Q(et.)-.11 E
(S_before is the sequence number of the last pack)129.6 402 Q
(et to arri)-.11 E .33 -.165(ve w)-.275 H(ith sequence number).165 E
(before S_loss.)129.6 415 Q
(S_after is the sequence number of the \214rst pack)129.6 431.6 Q
(et to arri)-.11 E .33 -.165(ve w)-.275 H(ith sequence number).165 E
(after S_loss.)129.6 444.6 Q(S_max is the lar)129.6 461.2 Q
(gest sequence number)-.198 E(.)-.605 E
(Therefore, S_before < S_loss < S_after <= S_max.)100.8 477.8 Q
(T_loss is the nominal estimated arri)129.6 494.4 Q -.275(va)-.275 G
2.75(lt).275 G(ime for the lost pack)-2.75 E(et.)-.11 E
(T_before is the reception time of S_before.)129.6 511 Q
(T_after is the reception time of S_after)129.6 527.6 Q(.)-.605 E(Note \
that T_before can either be before or after T_after due to reordering.)
100.8 544.2 Q -.165(Fo)100.8 560.8 S 2.75(ral).165 G(ost pack)-2.75 E
(et S_loss, we can interpolate its nominal "arri)-.11 E -.275(va)-.275 G
2.75(lt).275 G(ime" at the recei)-2.75 E -.165(ve)-.275 G 2.75(rf).165 G
(rom)-2.75 E(the arri)100.8 573.8 Q -.275(va)-.275 G 2.75(lt).275 G
(imes of S_before and S_after)-2.75 E 2.75(.T)-.605 G(hus:)-2.75 E
(T_loss = T_before + \( \(T_after - T_before\))136.8 592.8 Q 2.75(*\()
169.8 605.8 S(S_loss - S_before\)/\(S_after - S_before\) \);)-2.75 E
(Note that if the sequence space wrapped between S_before and S_after)
100.8 628.4 Q 2.75(,t)-.44 G(hen the sequence)-2.75 E
(numbers must be modi\214ed to tak)100.8 641.4 Q 2.75(et)-.11 G
(his into account before performing this calculation.)-2.75 E(If)5.5 E
(the lar)100.8 654.4 Q
(gest possible sequence number is S_max, and S_before > S_after)-.198 E
2.75(,t)-.44 G(hen modifying)-2.75 E(each sequence number S by S' = \(S\
 + \(S_max + 1\)/2\) mod \(S_max + 1\) w)100.8 667.4 Q(ould normally be)
-.11 E(suf)100.8 680.4 Q(\214cient.)-.275 E(Handle)72 733 Q(y/Flo)-.165
E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75
(5.2. [P)2.75 F(age 20])-.165 E 0 Cg EP
%%Page: 21 21
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(If the lost pack)100.8 85 Q
(et S_old w)-.11 E(as determined to ha)-.11 E .33 -.165(ve s)-.22 H
(tarted the pre).165 E(vious loss e)-.275 E -.165(ve)-.275 G
(nt, and we ha).165 E -.165(ve)-.22 G(just determined that S_ne)100.8 98
Q 2.75(wh)-.275 G(as been lost, then we interpolate the nominal arri)
-2.75 E -.275(va)-.275 G 2.75(lt).275 G(imes of)-2.75 E(S_old and S_ne)
100.8 111 Q 1.43 -.715(w, c)-.275 H(alled T_old and T_ne).715 E 2.75(wr)
-.275 G(especti)-2.75 E -.165(ve)-.275 G(ly).165 E(.)-.715 E
(If T_old + R >= T_ne)100.8 127.6 Q 1.43 -.715(w, t)-.275 H(hen S_ne)
.715 E 2.75(wi)-.275 G 2.75(sp)-2.75 G(art of the e)-2.75 E
(xisting loss e)-.165 E -.165(ve)-.275 G(nt. Otherwise S_ne).165 E 2.75
(wi)-.275 G(s)-2.75 E(the \214rst pack)100.8 140.6 Q(et in a ne)-.11 E
2.75(wl)-.275 G(oss e)-2.75 E -.165(ve)-.275 G(nt.).165 E/F1 11
/Times-Bold@0 SF(5.3.)72 170.2 Q/F2 13/Times-Bold@0 SF(Inter)5.5 E
(-loss Ev)-.481 E(ent Inter)-.13 E -.13(va)-.13 G(l).13 E F0
(If a loss interv)100.8 186.8 Q(al, A, is determined to ha)-.275 E .33
-.165(ve s)-.22 H(tarted with pack).165 E(et sequence number S_A and)
-.11 E(the ne)100.8 199.8 Q(xt loss interv)-.165 E
(al, B, started with pack)-.275 E
(et sequence number S_B, then the number of)-.11 E(pack)100.8 212.8 Q
(ets in loss interv)-.11 E(al A is gi)-.275 E -.165(ve)-.275 G 2.75(nb)
.165 G 2.75(y\()-2.75 G(S_B - S_A\).)-2.75 E(Thus, loss interv)5.5 E
(al A contains all of the)-.275 E(pack)100.8 225.8 Q
(ets transmitted by the sender starting with the \214rst pack)-.11 E
(et transmitted in loss interv)-.11 E(al)-.275 E(A, and ending with b)
100.8 238.8 Q(ut not including the \214rst pack)-.22 E
(et transmitted in loss interv)-.11 E(al B.)-.275 E F1(5.4.)72 277.8 Q
F2 -.78 -1.3(Av e)5.5 H(rage Loss Inter)1.3 E -.13(va)-.13 G(l).13 E F0
1.76 -.88(To c)100.8 294.4 T(alculate the loss e).88 E -.165(ve)-.275 G
(nt rate p, we \214rst calculate the a).165 E -.165(ve)-.22 G
(rage loss interv).165 E 2.75(al. This)-.275 F(is done)2.75 E
(using a \214lter that weights the n most recent loss e)100.8 307.4 Q
-.165(ve)-.275 G(nt interv).165 E(als in such a w)-.275 E(ay that the)
-.11 E(measured loss e)100.8 320.4 Q -.165(ve)-.275 G
(nt rate changes smoothly).165 E 5.5(.I)-.715 G 2.75(ft)-5.5 G(he recei)
-2.75 E -.165(ve)-.275 G 2.75(rh).165 G(as not yet seen a lost or)-2.75
E(mark)100.8 333.4 Q(ed pack)-.11 E(et, then the recei)-.11 E -.165(ve)
-.275 G 2.75(rd).165 G(oesn')-2.75 E 2.75(tc)-.198 G(alculate the a)
-2.75 E -.165(ve)-.22 G(rage loss interv).165 E(al.)-.275 E -.88(We)
100.8 350 S(ights w_0 to w_\(n-1\) are calculated as:).88 E
(If \(i < n/2\))136.8 369 Q(w_i = 1;)145.05 382 Q(Else)136.8 395 Q
(w_i = 1 - \(i - \(n/2 - 1\)\)/\(n/2 + 1\);)145.05 408 Q
(Thus if n=8, the v)100.8 430.6 Q(alues of w_0 to w_7 are:)-.275 E
(1.0, 1.0, 1.0, 1.0, 0.8, 0.6, 0.4, 0.2)129.6 447.2 Q(The v)100.8 463.8
Q(alue n for the number of loss interv)-.275 E
(als used in calculating the loss e)-.275 E -.165(ve)-.275 G(nt rate)
.165 E(determines TFRC')100.8 476.8 Q 2.75(ss)-.605 G
(peed in responding to changes in the le)-2.75 E -.165(ve)-.275 G 2.75
(lo).165 G 2.75(fc)-2.75 G 2.75(ongestion. As)-2.75 F(currently)2.75 E
(speci\214ed, TFRC SHOULD NO)100.8 489.8 Q 2.75(Tu)-.44 G(se v)-2.75 E
(alues of n signi\214cantly greater than 8, for traf)-.275 E(\214c that)
-.275 E(might compete in the global Internet with TCP)100.8 502.8 Q 5.5
(.A)-1.221 G 2.75(tt)-5.5 G(he v)-2.75 E(ery least, safe operation with)
-.165 E -.275(va)100.8 515.8 S(lues of n greater than 8 w).275 E
(ould require a slight change to TFRC')-.11 E 2.75(sm)-.605 G
(echanisms to include)-2.75 E 2.75(am)100.8 528.8 S(ore se)-2.75 E -.165
(ve)-.275 G(re response to tw).165 E 2.75(oo)-.11 G 2.75(rm)-2.75 G
(ore round-trip times with hea)-2.75 E(vy pack)-.22 E(et loss.)-.11 E
(When calculating the a)100.8 545.4 Q -.165(ve)-.22 G(rage loss interv)
.165 E(al we need to decide whether to include the current)-.275 E
(loss interv)100.8 558.4 Q(al, de\214ned as the loss interv)-.275 E
(al containing the most recent loss e)-.275 E -.165(ve)-.275 G 2.75
(nt. W).165 F 2.75(eo)-.88 G(nly)-2.75 E
(include the current loss interv)100.8 571.4 Q(al if it is suf)-.275 E
(\214ciently lar)-.275 E(ge to increase the a)-.198 E -.165(ve)-.22 G
(rage loss).165 E(interv)100.8 584.4 Q(al.)-.275 E
(Let the most recent loss interv)100.8 601 Q
(als be I_0 to I_k, where I_0 is the current loss interv)-.275 E 2.75
(al. If)-.275 F(there ha)100.8 614 Q .33 -.165(ve b)-.22 H
(een at least n loss interv).165 E
(als, then k is set to n; otherwise k is the maximum)-.275 E
(number of loss interv)100.8 627 Q(als seen so f)-.275 E(ar)-.11 E 5.5
(.W)-.605 G 2.75(ec)-6.38 G(alculate the a)-2.75 E -.165(ve)-.22 G
(rage loss interv).165 E(al I_mean as)-.275 E(follo)100.8 640 Q(ws:)
-.275 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055
E 189.593(idmer Section)-.44 F 2.75(5.4. [P)2.75 F(age 21])-.165 E 0 Cg
EP
%%Page: 22 22
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(I_tot0 = 0;)136.8 85 Q
(I_tot1 = 0;)136.8 98 Q(W_tot = 0;)136.8 111 Q(for \(i = 0 to k-1\) {)
136.8 124 Q(I_tot0 = I_tot0 + \(I_i * w_i\);)142.3 137 Q
(W_tot = W_tot + w_i;)142.3 150 Q(})136.8 163 Q(for \(i = 1 to k\) {)
136.8 176 Q(I_tot1 = I_tot1 + \(I_i * w_\(i-1\)\);)142.3 189 Q(})136.8
202 Q(I_tot = max\(I_tot0, I_tot1\);)136.8 215 Q(I_mean = I_tot/W_tot;)
136.8 228 Q(The loss e)100.8 247 Q -.165(ve)-.275 G
(nt rate, p is simply:).165 E 2.75(p=1/I)136.8 266 S(_mean;)-2.75 E/F1
11/Times-Bold@0 SF(5.5.)72 298 Q/F2 13/Times-Bold@0 SF
(History Discounting)5.5 E F0
(As described in Section 5.4, when there ha)100.8 314.6 Q .33 -.165
(ve b)-.22 H(een at least eight loss interv).165 E(als, the most)-.275 E
(recent loss interv)100.8 327.6 Q(al is only assigned 1/\(0.75*n\) of t\
he total weight in calculating the a)-.275 E -.165(ve)-.22 G(rage).165 E
(loss interv)100.8 340.6 Q(al, re)-.275 E -.055(ga)-.165 G
(rdless of the size of the most recent loss interv).055 E 2.75(al. This)
-.275 F(section describes)2.75 E(an optional history discounting mechan\
ism, discussed further in [FHPW00a] and [W00],)100.8 353.6 Q(that allo)
100.8 366.6 Q(ws the TFRC recei)-.275 E -.165(ve)-.275 G 2.75(rt).165 G
2.75(oa)-2.75 G(djust the weights, concentrating more of the relati)
-2.75 E -.165(ve)-.275 G(weight on the most recent loss interv)100.8
379.6 Q(al, when the most recent loss interv)-.275 E(al is more than)
-.275 E(twice as lar)100.8 392.6 Q(ge as the computed a)-.198 E -.165
(ve)-.22 G(rage loss interv).165 E(al.)-.275 E 1.76 -.88(To c)100.8
409.2 T(arry out history discounting, we associate a discount f).88 E
(actor DF_i with each loss)-.11 E(interv)100.8 422.2 Q
(al L_i, for i > 0, where each discount f)-.275 E
(actor is a \215oating point number)-.11 E 5.5(.T)-.605 G(he discount)
-5.5 E(array maintains the cumulati)100.8 435.2 Q .33 -.165(ve h)-.275 H
(istory of discounting for each loss interv).165 E 2.75(al. At)-.275 F
(the)2.75 E(be)100.8 448.2 Q(ginning, the v)-.165 E
(alues of DF_i in the discount array are initialized to 1:)-.275 E
(for \(i = 0 to n\) {)136.8 467.2 Q(DF_i = 1;)142.3 480.2 Q(})136.8
493.2 Q(History discounting also uses a general discount f)100.8 512.2 Q
(actor DF)-.11 E 2.75(,a)-.88 G(lso a \215oating point number)-2.75 E(,)
-.44 E(that is also initialized to 1.)100.8 525.2 Q(First we sho)5.5 E
2.75(wh)-.275 G .55 -.275(ow t)-2.75 H(he discount f).275 E
(actors are used in calculating)-.11 E(the a)100.8 538.2 Q -.165(ve)-.22
G(rage loss interv).165 E
(al, and then we describe later in this section ho)-.275 E 2.75(wt)-.275
G(he discount f)-2.75 E(actors)-.11 E(are modi\214ed o)100.8 551.2 Q
-.165(ve)-.165 G 2.75(rt).165 G(ime.)-2.75 E
(As described in Section 5.4 the a)100.8 567.8 Q -.165(ve)-.22 G
(rage loss interv).165 E(al is calculated using the n pre)-.275 E
(vious loss)-.275 E(interv)100.8 580.8 Q
(als I_1, ..., I_n and the current loss interv)-.275 E(al I_0.)-.275 E
(The computation of the a)5.5 E -.165(ve)-.22 G(rage loss).165 E(interv)
100.8 593.8 Q(al using the discount f)-.275 E
(actors is a simple modi\214cation of the procedure in Section 5.4,)-.11
E(as follo)100.8 606.8 Q(ws:)-.275 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)
-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75
(5.5. [P)2.75 F(age 22])-.165 E 0 Cg EP
%%Page: 23 23
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(I_tot0 = I_0 * w_0)136.8 85 Q
(I_tot1 = 0;)136.8 98 Q(W_tot0 = w_0)136.8 111 Q(W_tot1 = 0;)136.8 124 Q
(for \(i = 1 to n-1\) {)136.8 137 Q
(I_tot0 = I_tot0 + \(I_i * w_i * DF_i * DF\);)142.3 150 Q
(W_tot0 = W_tot0 + w_i * DF_i * DF;)142.3 163 Q(})136.8 176 Q
(for \(i = 1 to n\) {)136.8 189 Q
(I_tot1 = I_tot1 + \(I_i * w_\(i-1\) * DF_i\);)142.3 202 Q
(W_tot1 = W_tot1 + w_\(i-1\) * DF_i;)142.3 215 Q(})136.8 228 Q 2.75(p=m)
136.8 241 S(in\(W_tot0/I_tot0, W_tot1/I_tot1\);)-2.75 E
(The general discounting f)100.8 260 Q(actor DF is updated on e)-.11 E
-.165(ve)-.275 G(ry pack).165 E(et arri)-.11 E -.275(va)-.275 G 2.75(la)
.275 G 2.75(sf)-2.75 G(ollo)-2.75 E(ws. First, the)-.275 E(recei)100.8
273 Q -.165(ve)-.275 G 2.75(rc).165 G(omputes the weighted a)-2.75 E
-.165(ve)-.22 G(rage I_mean of the loss interv).165 E
(als I_1, ..., I_n:)-.275 E(I_tot = 0;)136.8 292 Q(W_tot = 0;)136.8 305
Q(for \(i = 1 to n\) {)136.8 318 Q(W_tot = W_tot + w_\(i-1\) * DF_i;)
142.3 331 Q(I_tot = I_tot + \(I_i * w_\(i-1\) * DF_i\);)142.3 344 Q(})
136.8 357 Q(I_mean = I_tot / W_tot;)136.8 370 Q(This weighted a)100.8
389 Q -.165(ve)-.22 G
(rage I_mean is compared to I_0, the size of current loss interv).165 E
2.75(al. If)-.275 F(I_0)2.75 E
(is greater than twice I_mean, then the ne)100.8 402 Q 2.75(wl)-.275 G
(oss interv)-2.75 E(al is considerably lar)-.275 E(ger than the old)
-.198 E(ones, and the general discount f)100.8 415 Q
(actor DF is updated to decrease the relati)-.11 E .33 -.165(ve w)-.275
H(eight on the).165 E(older interv)100.8 428 Q(als, as follo)-.275 E
(ws:)-.275 E(if \(I_0 > 2 * I_mean\) {)136.8 447 Q(DF = 2 * I_mean/I_0;)
142.3 460 Q(if \(DF < THRESHOLD\))142.3 473 Q(DF = THRESHOLD;)147.8 486
Q 2.75(}e)136.8 499 S(lse)-2.75 E(DF = 1;)142.3 512 Q 2.75(An)100.8 531
S(onzero v)-2.75 E(alue for THRESHOLD ensures that older loss interv)
-.275 E(als from an earlier time of)-.275 E
(high congestion are not discounted entirely)100.8 544 Q 5.5(.W)-.715 G
2.75(er)-6.38 G(ecommend a THRESHOLD of 0.25.)-2.75 E
(Note that with each ne)100.8 557 Q 2.75(wp)-.275 G(ack)-2.75 E(et arri)
-.11 E -.275(va)-.275 G(l, I_0 will increase further).275 E 2.75(,a)-.44
G(nd the discount f)-2.75 E(actor DF)-.11 E(will be updated.)100.8 570 Q
(When a ne)100.8 586.6 Q 2.75(wl)-.275 G(oss e)-2.75 E -.165(ve)-.275 G
(nt occurs, the current interv).165 E
(al shifts from I_0 to I_1, loss interv)-.275 E(al I_i)-.275 E
(shifts to interv)100.8 599.6 Q(al I_\(i+1\), and the loss interv)-.275
E(al I_n is for)-.275 E 2.75(gotten. The)-.198 F(pre)2.75 E
(vious discount f)-.275 E(actor)-.11 E
(DF has to be incorporated into the discount array)100.8 612.6 Q 5.5(.B)
-.715 G(ecause DF_i carries the discount f)-5.5 E(actor)-.11 E
(associated with loss interv)100.8 625.6 Q
(al I_i, the DF_i array has to be shifted as well. This is done as)-.275
E(follo)100.8 638.6 Q(ws:)-.275 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)
-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75
(5.5. [P)2.75 F(age 23])-.165 E 0 Cg EP
%%Page: 24 24
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(for \(i = 1 to n\) {)136.8 85 Q
(DF_i = DF * DF_i;)142.3 98 Q(})136.8 111 Q
(for \(i = n-1 to 0 step -1\) {)136.8 124 Q(DF_\(i+1\) = DF_i;)142.3 137
Q(})136.8 150 Q(I_0 = 1;)136.8 163 Q(DF_0 = 1;)136.8 176 Q(DF = 1;)136.8
189 Q(This completes the description of the optional history discountin\
g mechanism. W)100.8 211.6 Q(e)-.88 E(emphasize that this is an optiona\
l mechanism whose sole purpose is to allo)100.8 224.6 Q 2.75(wT)-.275 G
(FRC to)-2.75 E(response some)100.8 237.6 Q(what more quickly to the su\
dden absence of congestion, as represented by a)-.275 E
(long current loss interv)100.8 250.6 Q(al.)-.275 E/F1 11/Times-Bold@0
SF(6.)72 289.6 Q/F2 14/Times-Bold@0 SF(Data Recei)5.5 E -.14(ve)-.14 G
3.5(rP).14 G -.252(ro)-3.5 G(tocol).252 E F0(The recei)100.8 306.2 Q
-.165(ve)-.275 G 2.75(rp).165 G
(eriodically sends feedback messages to the sender)-2.75 E 5.5(.F)-.605
G(eedback pack)-5.5 E(ets should)-.11 E
(normally be sent at least once per R)100.8 319.2 Q(TT)-.66 E 2.75(,u)
-.814 G(nless the sender is sending at a rate of less than)-2.75 E
(one pack)100.8 332.2 Q(et per R)-.11 E(TT)-.66 E 2.75(,i)-.814 G 2.75
(nw)-2.75 G(hich case a feedback pack)-2.75 E(et should be send for e)
-.11 E -.165(ve)-.275 G(ry data pack).165 E(et)-.11 E(recei)100.8 345.2
Q -.165(ve)-.275 G 2.75(d. A).165 F(feedback pack)2.75 E
(et should also be sent whene)-.11 E -.165(ve)-.275 G 2.75(ran).165 G
.55 -.275(ew l)-2.75 H(oss e).275 E -.165(ve)-.275 G(nt is detected).165
E(without w)100.8 358.2 Q(aiting for the end of an R)-.11 E(TT)-.66 E
2.75(,a)-.814 G(nd whene)-2.75 E -.165(ve)-.275 G 2.75(ra).165 G 2.75
(no)-2.75 G(ut-of-order data pack)-2.75 E(et is)-.11 E(recei)100.8 371.2
Q -.165(ve)-.275 G 2.75(dt).165 G(hat remo)-2.75 E -.165(ve)-.165 G 2.75
(sal).165 G(oss e)-2.75 E -.165(ve)-.275 G(nt from the history).165 E(.)
-.715 E(If the sender is transmitting at a high rate \(man)100.8 387.8 Q
2.75(yp)-.165 G(ack)-2.75 E(ets per R)-.11 E(TT\) there may be some)-.66
E(adv)100.8 400.8 Q
(antages to sending periodic feedback messages more than once per R)
-.275 E(TT as this allo)-.66 E(ws)-.275 E -.11(fa)100.8 413.8 S
(ster response to changing R).11 E
(TT measurements, and more resilience to feedback pack)-.66 E(et)-.11 E
2.75(loss. If)100.8 426.8 R(the recei)2.75 E -.165(ve)-.275 G 2.75(rw)
.165 G(as sending k feedback pack)-2.86 E(ets per R)-.11 E(TT)-.66 E
2.75(,s)-.814 G(tep \(4\) of Section 6.2 w)-2.75 E(ould)-.11 E
(be modi\214ed to set the feedback timer to e)100.8 439.8 Q
(xpire after R_m/k seconds.)-.165 E(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G
.88 -.44(r, e).165 H(ach).44 E(feedback pack)100.8 452.8 Q(et w)-.11 E
(ould still report the recei)-.11 E -.165(ve)-.275 G 2.75(rr).165 G
(ate o)-2.75 E -.165(ve)-.165 G 2.75(rt).165 G(he last R)-2.75 E(TT)-.66
E 2.75(,n)-.814 G(ot o)-2.75 E -.165(ve)-.165 G 2.75(raf).165 G
(raction of)-2.75 E(an R)100.8 465.8 Q(TT)-.66 E 5.5(.W)-.814 G 2.75(en)
-6.38 G(ote that there is little g)-2.75 E(ain from sending a lar)-.055
E(ge number of feedback)-.198 E(messages per R)100.8 478.8 Q(TT)-.66 E
(.)-.814 E F1(6.1.)72 517.8 Q/F3 13/Times-Bold@0 SF(Recei)5.5 E -.13(ve)
-.13 G 3.25(rB).13 G(eha)-3.25 E(vior When a Data P)-.325 E(ack)-.13 E
(et is Recei)-.13 E -.13(ve)-.13 G(d).13 E F0(When a data pack)100.8
534.4 Q(et is recei)-.11 E -.165(ve)-.275 G(d, the recei).165 E -.165
(ve)-.275 G 2.75(rp).165 G(erforms the follo)-2.75 E(wing steps:)-.275 E
16.887(1\) Add)100.8 551 R(the pack)2.75 E(et to the pack)-.11 E
(et history)-.11 E(.)-.715 E 16.887(2\) Check)100.8 567.6 R
(if done: If the ne)2.75 E 2.75(wp)-.275 G(ack)-2.75 E
(et results in the detection of a ne)-.11 E 2.75(wl)-.275 G(oss e)-2.75
E -.165(ve)-.275 G(nt, or if no).165 E(feedback pack)129.6 580.6 Q(et w)
-.11 E(as sent when the feedback timer last e)-.11 E
(xpired, go to step 3\).)-.165 E(Otherwise, no action need be performed\
 \(unless the optimization in the ne)129.6 593.6 Q(xt)-.165 E
(paragraph is used\), so e)129.6 606.6 Q(xit the procedure.)-.165 E
(An optimization might check to see if the arri)129.6 623.2 Q -.275(va)
-.275 G 2.75(lo).275 G 2.75(ft)-2.75 G(he pack)-2.75 E
(et caused a hole in the)-.11 E(pack)129.6 636.2 Q
(et history to be \214lled and consequently tw)-.11 E 2.75(ol)-.11 G
(oss interv)-2.75 E(als were mer)-.275 E(ged into one.)-.198 E
(If this is the case, the recei)129.6 649.2 Q -.165(ve)-.275 G 2.75(rm)
.165 G(ight also send feedback immediately)-2.75 E 5.5(.T)-.715 G(he ef)
-5.5 E(fects of)-.275 E(such an optimization are normally e)129.6 662.2
Q(xpected to be small.)-.165 E 16.887(3\) Calculate)100.8 678.8 R
(p: Let the pre)2.75 E(vious v)-.275 E(alue of p be p_pre)-.275 E 4.18
-.715(v. C)-.275 H(alculate the ne).715 E 2.75(wv)-.275 G(alue of p as)
-3.025 E(described in Section 5.)129.6 691.8 Q(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44
F 2.75(6.1. [P)2.75 F(age 24])-.165 E 0 Cg EP
%%Page: 25 25
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 16.887(4\) Expire)100.8 85 R
(feedback timer?: If p > p_pre)2.75 E 1.43 -.715(v, c)-.275 H
(ause the feedback timer to e).715 E(xpire, and)-.165 E
(perform the actions described in Section 6.2)129.6 98 Q(If p <= p_pre)
129.6 114.6 Q 2.75(va)-.275 G(nd no feedback pack)-2.75 E(et w)-.11 E
(as sent when the feedback timer last)-.11 E -.165(ex)129.6 127.6 S
(pired, cause the feedback timer to e).165 E
(xpire, and perform the actions described in)-.165 E
(Section 6.2 If p <= p_pre)129.6 140.6 Q 2.75(va)-.275 G
(nd a feedback pack)-2.75 E(et w)-.11 E(as sent when the feedback timer)
-.11 E(last e)129.6 153.6 Q(xpired, no action need be performed.)-.165 E
/F1 11/Times-Bold@0 SF(6.2.)72 196.2 Q/F2 13/Times-Bold@0 SF
(Expiration of F)5.5 E(eedback T)-.325 E(imer)-.234 E F0
(When the feedback timer at the recei)100.8 212.8 Q -.165(ve)-.275 G
2.75(re).165 G(xpires, the action to be tak)-2.915 E
(en depends on whether)-.11 E(data pack)100.8 225.8 Q(ets ha)-.11 E .33
-.165(ve b)-.22 H(een recei).165 E -.165(ve)-.275 G 2.75(ds).165 G
(ince the last feedback w)-2.75 E(as sent.)-.11 E -.165(Fo)100.8 242.4 S
2.75(rt).165 G(he m-th e)-2.75 E(xpiration of the feedback timer)-.165 E
2.75(,l)-.44 G(et the maximum sequence number of a)-2.75 E(pack)100.8
255.4 Q(et at the recei)-.11 E -.165(ve)-.275 G 2.75(rs).165 G 2.75(of)
-2.75 G(ar be S_m, and the v)-2.86 E(alue of the R)-.275 E
(TT measurement included in)-.66 E(pack)100.8 268.4 Q(et S_m be R_m.)
-.11 E(As described in Section 3.2.1, R_m is the sender')5.5 E 2.75(sc)
-.605 G(urrent estimate of)-2.75 E
(the round trip time, reported in data pack)100.8 281.4 Q 2.75(ets. If)
-.11 F(data pack)2.75 E(ets ha)-.11 E .33 -.165(ve b)-.22 H(een recei)
.165 E -.165(ve)-.275 G 2.75(ds).165 G(ince the)-2.75 E(pre)100.8 294.4
Q(vious feedback w)-.275 E(as sent, the recei)-.11 E -.165(ve)-.275 G
2.75(rp).165 G(erforms the follo)-2.75 E(wing steps:)-.275 E 16.887
(1\) Calculate)100.8 311 R(the a)2.75 E -.165(ve)-.22 G(rage loss e).165
E -.165(ve)-.275 G(nt rate using the algorithm described abo).165 E
-.165(ve)-.165 G(.).165 E 12.003(2a\) If)100.8 327.6 R
(the feedback timer e)2.75 E(xpired at its normal time, or e)-.165 E
(xpired early due to a ne)-.165 E 2.75(wl)-.275 G(ost or)-2.75 E(mark)
129.6 340.6 Q(ed pack)-.11 E
(et \(i.e., step \(3\) in Section 6.1\), calculate the measured recei)
-.11 E .33 -.165(ve r)-.275 H(ate,).165 E(X_recv)129.6 353.6 Q 2.75(,b)
-.715 G(ased on the pack)-2.75 E(ets recei)-.11 E -.165(ve)-.275 G 2.75
(dw).165 G(ithin the pre)-2.75 E(vious R_\(m-1\) seconds.)-.275 E
(In the)5.5 E(typical case, when the recei)129.6 366.6 Q -.165(ve)-.275
G 2.75(ri).165 G 2.75(ss)-2.75 G(ending only one feedback pack)-2.75 E
(et per round-trip)-.11 E(time and the feedback timer did not e)129.6
379.6 Q(xpire early due to an idle period, then R_\(m-1\))-.165 E -.11
(wo)129.6 392.6 S(uld be the time interv).11 E
(al since the feedback timer last e)-.275 E(xpired.)-.165 E 11.387
(2b\) If)100.8 409.2 R(the feedback timer e)2.75 E
(xpired early due to a ne)-.165 E 2.75(wl)-.275 G(ost or mark)-2.75 E
(ed pack)-.11 E(et \(i.e., step \(3\))-.11 E
(in Section 6.1\), the recei)129.6 422.2 Q .33 -.165(ve r)-.275 H
(ate X_recv SHOULD be calculated based on the pack).165 E(ets)-.11 E
(recei)129.6 435.2 Q -.165(ve)-.275 G 2.75(dw).165 G(ithin the pre)-2.75
E(vious R_\(m-1\) seconds.)-.275 E -.165(Fo)5.5 G 2.75(re).165 G
(ase of implementation, the)-2.75 E(recei)129.6 448.2 Q .33 -.165(ve r)
-.275 H(ate MA).165 E 2.75(Yb)-1.155 G 2.75(ec)-2.75 G(alculated o)-2.75
E -.165(ve)-.165 G 2.75(ral).165 G(onger time interv)-2.75 E
(al, the time interv)-.275 E(al going)-.275 E
(back to the most recent feedback timer e)129.6 461.2 Q
(xpiration that w)-.165 E(as at least R_\(m-1\) seconds)-.11 E(ago.)
129.6 474.2 Q 16.887(3\) Prepare)100.8 490.8 R(and send a feedback pack)
2.75 E(et containing the information described in Section)-.11 E(3.2.2.)
129.6 503.8 Q 16.887(4\) Restart)100.8 520.4 R(the feedback timer to e)
2.75 E(xpire after R_m seconds.)-.165 E(Note that rule 2\) abo)100.8 537
Q .33 -.165(ve g)-.165 H -2.365 -.275(iv e).165 H 2.75(sam).275 G
(inimum v)-2.75 E(alue for the measured recei)-.275 E .33 -.165(ve r)
-.275 H(ate X_recv of one).165 E(pack)100.8 550 Q
(et per round-trip time.)-.11 E
(If the sender is limited to a sending rate of less than one pack)5.5 E
(et)-.11 E(per round-trip time, this will be due to the loss e)100.8 563
Q -.165(ve)-.275 G(nt rate, not from a limit imposed by the).165 E
(measured recei)100.8 576 Q .33 -.165(ve r)-.275 H(ate at the recei).165
E -.165(ve)-.275 G -.605(r.).165 G(If no data pack)100.8 592.6 Q(ets ha)
-.11 E .33 -.165(ve b)-.22 H(een recei).165 E -.165(ve)-.275 G 2.75(ds)
.165 G(ince the last feedback w)-2.75 E(as sent, no feedback pack)-.11 E
(et)-.11 E(is sent, and the feedback timer is restarted to e)100.8 605.6
Q(xpire after R_m seconds.)-.165 E F1(6.3.)72 648.2 Q F2(Recei)5.5 E
-.13(ve)-.13 G 3.25(rI).13 G(nitialization)-3.25 E F0(The recei)100.8
664.8 Q -.165(ve)-.275 G 2.75(ri).165 G 2.75(si)-2.75 G
(nitialized by the \214rst data pack)-2.75 E(et that arri)-.11 E -.165
(ve)-.275 G 2.75(sa).165 G 2.75(tt)-2.75 G(he recei)-2.75 E -.165(ve)
-.275 G 1.21 -.605(r. L).165 H(et the).605 E
(sequence number of this pack)100.8 677.8 Q(et be i.)-.11 E(Handle)72
733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593
(idmer Section)-.44 F 2.75(6.3. [P)2.75 F(age 25])-.165 E 0 Cg EP
%%Page: 26 26
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(When the \214rst pack)100.8 85
Q(et is recei)-.11 E -.165(ve)-.275 G(d:).165 E 24.95<8353>100.8 101.6 S
(et p=0)-24.95 E 24.95<8353>100.8 118.2 S 2.75(et X_recv)-24.95 F 2.75
(=0)2.75 G(.)-2.75 E 24.95<8350>100.8 134.8 S
(repare and send a feedback pack)-24.95 E(et.)-.11 E 24.95<8353>100.8
151.4 S(et the feedback timer to e)-24.95 E(xpire after R_i seconds.)
-.165 E(If the \214rst data pack)100.8 168 Q(et doesn')-.11 E 2.75(tc)
-.198 G(ontain an estimate R_i of the round-trip time, then the)-2.75 E
(recei)100.8 181 Q -.165(ve)-.275 G 2.75(rs).165 G(ends a feedback pack)
-2.75 E(et for e)-.11 E -.165(ve)-.275 G(ry arri).165 E(ving data pack)
-.275 E(et, until a data pack)-.11 E(et arri)-.11 E -.165(ve)-.275 G(s)
.165 E(containing an estimate of the round-trip time.)100.8 194 Q
(If the sender is using a coarse-grained timestamp that increments e)
100.8 210.6 Q -.165(ve)-.275 G(ry quarter of a round-).165 E
(trip time, then a feedback timer is not needed, and the follo)100.8
223.6 Q(wing procedure from RFC 4342)-.275 E
(is used to determine when to send feedback messages.)100.8 236.6 Q
24.95<8357>100.8 253.2 S(hene)-24.95 E -.165(ve)-.275 G 2.75(rt).165 G
(he recei)-2.75 E -.165(ve)-.275 G 2.75(rs).165 G
(ends a feedback message, the recei)-2.75 E -.165(ve)-.275 G 2.75(rs)
.165 G(ets a local v)-2.75 E(ariable)-.275 E
(last_counter to the greatest recei)129.6 266.2 Q -.165(ve)-.275 G 2.75
(dv).165 G(alue of the windo)-3.025 E 2.75(wc)-.275 G
(ounter since the last)-2.75 E(feedback message w)129.6 279.2 Q
(as sent, if an)-.11 E 2.75(yd)-.165 G(ata pack)-2.75 E(ets ha)-.11 E
.33 -.165(ve b)-.22 H(een recei).165 E -.165(ve)-.275 G 2.75(ds).165 G
(ince the last)-2.75 E(feedback message w)129.6 292.2 Q(as sent.)-.11 E
24.95<8349>100.8 308.8 S 2.75(ft)-24.95 G(he recei)-2.75 E -.165(ve)
-.275 G 2.75(rr).165 G(ecei)-2.75 E -.165(ve)-.275 G 2.75(sad).165 G
(ata pack)-2.75 E(et with a windo)-.11 E 2.75(wc)-.275 G(ounter v)-2.75
E(alue greater than or)-.275 E
(equal to last_counter + 4, then the recei)129.6 321.8 Q -.165(ve)-.275
G 2.75(rs).165 G(ends a ne)-2.75 E 2.75(wf)-.275 G(eedback pack)-2.75 E
2.75(et. \("Greater")-.11 F
(and "greatest" are measured in circular windo)129.6 334.8 Q 2.75(wc)
-.275 G(ounter space.\))-2.75 E/F1 11/Times-Bold@0 SF 2.75
(6.3.1. Initializing)72 360.8 R
(the Loss History after the First Loss Ev)2.75 E(ent)-.11 E F0
(The number of pack)100.8 377.4 Q
(ets until the \214rst loss can not be used to compute the allo)-.11 E
(wed sending)-.275 E(rate directly)100.8 390.4 Q 2.75(,a)-.715 G 2.75
(st)-2.75 G(he sending rate changes rapidly during this time.)-2.75 E
(TFRC assumes that the)5.5 E(correct data rate after the \214rst loss i\
s half of the maximum sending rate before the loss)100.8 403.4 Q 2.75
(occurred. TFRC)100.8 416.4 R(approximates this tar)2.75 E
(get rate X_tar)-.198 E(get by the maximum X_rec so f)-.198 E(ar)-.11 E
2.75(,f)-.44 G(or)-2.75 E(X_recv the recei)100.8 429.4 Q .33 -.165(ve r)
-.275 H(ate o).165 E -.165(ve)-.165 G 2.75(ras).165 G
(ingle round-trip time.)-2.75 E(\(F)5.5 E(or a TFRC sender that al)-.165
E -.11(wa)-.11 G(ys has).11 E(data to send, it is suf)100.8 442.4 Q
(\214cient to approximate the tar)-.275 E
(get rate by the most recent X_recv)-.198 E(.)-.715 E(Ho)100.8 455.4 Q
(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H(or a TFRC sender that\
 is sometimes data-limited or idle, it is best to use the).44 E
(maximum X_recv so f)100.8 468.4 Q(ar)-.11 E(.\))-.605 E(After the \214\
rst loss, instead of initializing the \214rst loss interv)100.8 494.4 Q
(al to the number of pack)-.275 E(ets sent)-.11 E
(until the \214rst loss, the TFRC recei)100.8 507.4 Q -.165(ve)-.275 G
2.75(rc).165 G(alculates the loss interv)-2.75 E(al that w)-.275 E
(ould be required to)-.11 E(produce the data rate X_tar)100.8 520.4 Q
(get, and uses this synthetic loss interv)-.198 E
(al to seed the loss history)-.275 E(mechanism.)100.8 533.4 Q
(TFRC does this by \214nding some v)100.8 559.4 Q
(alue p for which the throughput equation in Section 3.1)-.275 E(gi)
100.8 572.4 Q -.165(ve)-.275 G 2.75(sas).165 G
(ending rate within 5% of X_tar)-2.75 E(get, gi)-.198 E -.165(ve)-.275 G
2.75(nt).165 G(he round-trip time R, and the \214rst loss)-2.75 E
(interv)100.8 585.4 Q(al is then set to 1/p.)-.275 E(If the recei)5.5 E
-.165(ve)-.275 G 2.75(rk).165 G(no)-2.75 E(ws the se)-.275 E
(gment size s used by the sender)-.165 E 2.75(,t)-.44 G(hen)-2.75 E
(the recei)100.8 598.4 Q -.165(ve)-.275 G 2.75(rc).165 G
(an use the throughput equation for X; otherwise, the recei)-2.75 E
-.165(ve)-.275 G 2.75(rc).165 G(an measure the)-2.75 E(recei)100.8 611.4
Q .33 -.165(ve r)-.275 H(ate in pack).165 E(ets per second instead of b\
ytes per second for this purpose, and use the)-.11 E
(throughput equation for X_pps.)100.8 624.4 Q
(\(The 5% tolerance is introduced simply because the)5.5 E
(throughput equation is dif)100.8 637.4 Q(\214cult to in)-.275 E -.165
(ve)-.44 G(rt, and we w).165 E(ant to reduce the costs of calculating p)
-.11 E(numerically)100.8 650.4 Q(.\))-.715 E
(Special care is needed for initializing the \214rst loss interv)100.8
676.4 Q(al when the \214rst data pack)-.275 E(et is lost)-.11 E(or mark)
100.8 689.4 Q 2.75(ed. When)-.11 F(the \214rst data pack)2.75 E
(et is lost in TCP)-.11 E 2.75(,t)-1.221 G
(he TCP sender retransmits the pack)-2.75 E(et)-.11 E(Handle)72 733 Q
(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 181.343
(idmer Section)-.44 F 2.75(6.3.1. [P)2.75 F(age 26])-.165 E 0 Cg EP
%%Page: 27 27
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(after the retransmit timer e)
100.8 85 Q 2.75(xpires. If)-.165 F(TCP')2.75 E 2.75<738c>-.605 G
(rst data pack)-2.75 E(et is ECN-mark)-.11 E(ed, the TCP)-.11 E
(sender resets the retransmit timer)100.8 98 Q 2.75(,a)-.44 G
(nd sends a ne)-2.75 E 2.75(wd)-.275 G(ata pack)-2.75 E
(et only when the retransmit)-.11 E(timer e)100.8 111 Q
(xpires [RFC3168] \(Section 6.1.2\).)-.165 E -.165(Fo)5.5 G 2.75(rT).165
G(FRC, if the \214rst data pack)-2.75 E(et is lost or ECN-)-.11 E(mark)
100.8 124 Q(ed, then the \214rst loss interv)-.11 E
(al consists of the null interv)-.275 E(al with no data pack)-.275 E
2.75(ets. In)-.11 F(this)2.75 E(case, the loss interv)100.8 137 Q
(al length for this \(null\) loss interv)-.275 E(al should be set to gi)
-.275 E .33 -.165(ve a s)-.275 H(imilar).165 E
(sending rate to that of TCP)100.8 150 Q(.)-1.221 E
(When the \214rst TFRC loss interv)100.8 176 Q
(al is null, meaning that the \214rst data pack)-.275 E
(et is lost or ECN-)-.11 E(mark)100.8 189 Q(ed, in order to follo)-.11 E
2.75(wt)-.275 G(he beha)-2.75 E(vior of TCP)-.22 E 2.75(,T)-1.221 G
(FRC w)-2.75 E(ants the allo)-.11 E(wed sending rate to be)-.275 E 2.75
(1p)100.8 202 S(ack)-2.75 E(et e)-.11 E -.165(ve)-.275 G(ry tw).165 E
2.75(or)-.11 G(ound-trip times, or equi)-2.75 E -.275(va)-.275 G(lently)
.275 E 2.75(,0)-.715 G(.5 pack)-2.75 E(ets per R)-.11 E(TT)-.66 E 5.5
(.T)-.814 G(hus, the TFRC)-5.5 E(recei)100.8 215 Q -.165(ve)-.275 G 2.75
(rc).165 G(alculates the loss interv)-2.75 E(al that w)-.275 E
(ould be required to produce the tar)-.11 E(get rate)-.198 E(X_tar)100.8
228 Q(get of 0.5/R pack)-.198 E(ets per second, for the round-trip time\
 R, and uses this synthetic loss)-.11 E(interv)100.8 241 Q
(al for the \214rst loss interv)-.275 E 2.75(al. The)-.275 F(TFRC recei)
2.75 E -.165(ve)-.275 G 2.75(ru).165 G(ses 0.5/R pack)-2.75 E
(ets per second as the)-.11 E(minimum v)100.8 254 Q(alue for X_tar)-.275
E(get when initializing the \214rst loss interv)-.198 E(al.)-.275 E/F1
11/Times-Bold@0 SF(7.)72 293 Q/F2 14/Times-Bold@0 SF(Sender)5.5 E
(-based V)-.518 E(ariants)-1.288 E F0(In a sender)100.8 309.6 Q
(-based v)-.22 E(ariant of TFRC, the recei)-.275 E -.165(ve)-.275 G 2.75
(ru).165 G(ses reliable deli)-2.75 E -.165(ve)-.275 G
(ry to send information).165 E(about pack)100.8 322.6 Q
(et losses to the sender)-.11 E 2.75(,a)-.44 G
(nd the sender computes the pack)-2.75 E(et loss rate and the)-.11 E
(acceptable transmit rate.)100.8 335.6 Q(The main adv)100.8 361.6 Q
(antage of a sender)-.275 E(-based v)-.22 E
(ariant of TFRC is that the sender does not ha)-.275 E .33 -.165(ve t)
-.22 H(o).165 E(trust the recei)100.8 374.6 Q -.165(ve)-.275 G(r').165 E
2.75(sc)-.605 G(alculation of the pack)-2.75 E(et loss rate.)-.11 E(Ho)
5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, w).165 H
(ith the requirement of).44 E(reliable deli)100.8 387.6 Q -.165(ve)-.275
G(ry of loss information from the recei).165 E -.165(ve)-.275 G 2.75(rt)
.165 G 2.75(ot)-2.75 G(he sender)-2.75 E 2.75(,as)-.44 G(ender)-2.75 E
(-based TFRC)-.22 E -.11(wo)100.8 400.6 S(uld ha).11 E .33 -.165(ve m)
-.22 H(uch tighter constraints on the transport protocol in which it is\
 embedded.).165 E(In contrast, the recei)100.8 426.6 Q -.165(ve)-.275 G
-.22(r-).165 G(based v).22 E
(ariant of TFRC speci\214ed in this document is rob)-.275 E(ust to the)
-.22 E(loss of feedback pack)100.8 439.6 Q
(ets, and therefore does not require the reliable deli)-.11 E -.165(ve)
-.275 G(ry of feedback).165 E(pack)100.8 452.6 Q 2.75(ets. It)-.11 F
(is also better suited for applications where it is desirable to of)2.75
E(\215oad w)-.275 E(ork from)-.11 E(the serv)100.8 465.6 Q
(er to the client as much as possible.)-.165 E(RFC 4340 and RFC 4342 to\
gether specify CCID 3, which can be used as a sender)100.8 491.6 Q
(-based)-.22 E -.275(va)100.8 504.6 S(riant of TFRC.).275 E
(In CCID 3, each feedback pack)5.5 E(et from the recei)-.11 E -.165(ve)
-.275 G 2.75(rc).165 G(ontains a Loss)-2.75 E(Interv)100.8 517.6 Q
(als option, reporting the lengths of the most recent loss interv)-.275
E 2.75(als. Feedback)-.275 F(pack)2.75 E(ets)-.11 E
(may also include the Ack V)100.8 530.6 Q(ector option, allo)-1.221 E
(wing the sender to determine e)-.275 E(xactly which)-.165 E(pack)100.8
543.6 Q(ets were dropped or mark)-.11 E
(ed and to check the information reported in the Loss)-.11 E(Interv)
100.8 556.6 Q(als options.)-.275 E(The Ack V)5.5 E
(ector option can also include ECN Nonce Echoes, allo)-1.221 E(wing)
-.275 E(the sender to v)100.8 569.6 Q(erify the recei)-.165 E -.165(ve)
-.275 G(r').165 E 2.75(sr)-.605 G(eport of ha)-2.75 E(ving recei)-.22 E
-.165(ve)-.275 G 2.75(da).165 G 2.75(nu)-2.75 G(nmark)-2.75 E
(ed data pack)-.11 E 2.75(et. The)-.11 F(Ack V)100.8 582.6 Q
(ector option allo)-1.221 E
(ws the sender to see for itself which data pack)-.275 E
(ets were lost or ECN-)-.11 E(mark)100.8 595.6 Q
(ed, to determine loss interv)-.11 E(als, and to calculate the loss e)
-.275 E -.165(ve)-.275 G(nt rate.).165 E(Section 9.2 of RFC)5.5 E
(4342 discusses issues in the sender v)100.8 608.6 Q
(erifying information reported by the recei)-.165 E -.165(ve)-.275 G
-.605(r.).165 G F1(8.)72 647.6 Q F2(Implementation Issues)5.5 E F0(This\
 document has speci\214ed the TFRC congestion control mechanism, for us\
e by)100.8 664.2 Q(applications and transport protocols.)100.8 677.2 Q
(This section mentions brie\215y some of the)5.5 E
(implementation issues.)100.8 690.2 Q(Handle)72 733 Q(y/Flo)-.165 E
(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 197.843(idmer Section)-.44 F 2.75
(8. [P)2.75 F(age 27])-.165 E 0 Cg EP
%%Page: 28 28
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(Computing the throughput equation \(Section 3.1\):)100.8 85 Q -.165(Fo)
5.5 G 2.75(rt).165 G(_R)-2.75 E .396 -.198(TO = 4)-.66 H
(*R and b = 1, the).198 E(throughput equation in Section 3.1 can be e)
100.8 98 Q(xpressed as follo)-.165 E(ws:)-.275 E(s)169.8 117 Q(X_Bps =)
136.8 130 Q(--------)5.5 E 2.75(R*f)161.55 143 S(\(p\))-2.75 E(for)100.8
162 Q(f\(p\) =)136.8 181 Q
(sqrt\(2*p/3\) + \(12*sqrt\(3*p/8\) * p * \(1+32*p^2\)\).)5.5 E 2.75(At)
100.8 200 S(able lookup could be used for the function f\(p\).)-2.75 E
(Man)100.8 226 Q 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he multiplications \(e.g., q and 1-q for the round-trip time a)-2.75 E
-.165(ve)-.22 G(rage, a f).165 E(actor of 4)-.11 E
(for the timeout interv)100.8 239 Q(al\) are or could be by po)-.275 E
(wers of tw)-.275 E(o, and therefore could be)-.11 E
(implemented as simple shift operations.)100.8 252 Q
(The sender mechanism for pre)100.8 268.6 Q -.165(ve)-.275 G
(nting oscillations \(Section 4.5\):).165 E 1.76 -.88(We n)5.5 H
(ote that the optional).88 E(sender mechanism for pre)100.8 281.6 Q
-.165(ve)-.275 G
(nting oscillations described in Section 4.5 uses a square-root).165 E
(computation.)100.8 294.6 Q(Calculating the nominal pack)100.8 320.6 Q
(et arri)-.11 E -.275(va)-.275 G 2.75(lt).275 G(ime \(Section 5.2\).)
-2.75 E -.165(Fo)5.5 G 2.75(rt).165 G(he calculation of the)-2.75 E
(nominal arri)100.8 333.6 Q -.275(va)-.275 G 2.75(lt).275 G
(ime T_loss for a lost pack)-2.75 E(et from Section 5.2, one w)-.11 E
(ay to implement this)-.11 E(that w)100.8 346.6 Q(ould a)-.11 E -.22(vo)
-.22 G(id concerns about wrapped sequence space w).22 E
(ould be to use the follo)-.11 E(wing:)-.275 E(T_loss = T_before +)136.8
365.6 Q(\(T_after - T_before\))5.5 E 2.75(*D)150.55 378.6 S
(ist\(S_loss, S_before\)/Dist\(S_after)-2.75 E 2.75(,S)-.44 G(_before\))
-2.75 E(where)100.8 397.6 Q
(Dist\(Seqno_A, Seqno_B\) = \(Seqno_A + 2^48 - Seqno_B\) % 2^48)136.8
416.6 Q(The calculation of the a)100.8 448.6 Q -.165(ve)-.22 G
(rage loss interv).165 E(al \(Section 5.4\):)-.275 E
(The calculation of the a)5.5 E -.165(ve)-.22 G(rage).165 E(loss interv)
100.8 461.6 Q(al in Section 5.4 in)-.275 E -.22(vo)-.44 G(lv).22 E
(es multiplications by the weights w_0 to w_\(n-1\), which)-.165 E
(for n=8 are:)100.8 474.6 Q(1.0, 1.0, 1.0, 1.0, 0.8, 0.6, 0.4, 0.2.)
129.6 491.2 Q -.44(Wi)100.8 507.8 S(th a minor loss of smoothness, it w)
.44 E(ould be possible to use weights that were po)-.11 E(wers of)-.275
E(tw)100.8 520.8 Q 2.75(oo)-.11 G 2.75(rs)-2.75 G(ums of po)-2.75 E
(wers of tw)-.275 E(o, e.g.,)-.11 E
(1.0, 1.0, 1.0, 1.0, 0.75, 0.5, 0.25, 0.25.)129.6 537.4 Q
(The optional history discounting mechanism \(Section 5.5\):)100.8 554 Q
(The optional history)5.5 E(discounting mechanism described in Section \
5.5 is used in the calculation of the a)100.8 567 Q -.165(ve)-.22 G
(rage).165 E(loss rate.)100.8 580 Q
(The history discounting mechanism is in)5.5 E -.22(vo)-.44 G -.11(ke)
.22 G 2.75(do).11 G(nly when there has been an)-2.75 E
(unusually long interv)100.8 593 Q(al with no pack)-.275 E(et losses.)
-.11 E -.165(Fo)5.5 G 2.75(ram).165 G(ore ef)-2.75 E
(\214cient operation, the discount)-.275 E -.11(fa)100.8 606 S
(ctor DF_i could be restricted to be a po).11 E(wer of tw)-.275 E(o.)
-.11 E/F1 11/Times-Bold@0 SF(9.)72 645 Q/F2 14/Times-Bold@0 SF
(Changes fr)5.5 E(om RFC 3448)-.252 E F0
(This section summarizes the changes from RFC 3448.)100.8 661.6 Q
(Section 4.1, estimating the a)100.8 687.6 Q -.165(ve)-.22 G(rage se)
.165 E(gment size: Section 4.1 w)-.165 E(as modi\214ed to gi)-.11 E .33
-.165(ve a)-.275 H(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E
(ye/W)-.055 E 197.843(idmer Section)-.44 F 2.75(9. [P)2.75 F(age 28])
-.165 E 0 Cg EP
%%Page: 29 29
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(speci\214c algorithm that could be used for estimating the a)100.8 85 Q
-.165(ve)-.22 G(rage se).165 E(gment size.)-.165 E(Section 4.2, update \
to the initial sending rate: In RFC 3448, the initial sending rate w)
100.8 111 Q(as tw)-.11 E(o)-.11 E(pack)100.8 124 Q
(ets per round trip time.)-.11 E
(In this document, the initial sending rate can be as high as four)5.5 E
(pack)100.8 137 Q(ets per round trip time, follo)-.11 E(wing RFC 3390.)
-.275 E(The initial sending rate w)5.5 E(as changed to)-.11 E
(be in terms of the se)100.8 150 Q
(gment size s, not in terms of the MSS.)-.165 E(Section 4.2 no)100.8 176
Q 2.75(ws)-.275 G(ays that tld, the T)-2.75 E
(ime Last Doubled during slo)-.385 E(w-start, can be initialized to)
-.275 E(either 0 or to -1.)100.8 189 Q(Section 4.2 w)5.5 E
(as also clari\214ed to say that R)-.11 E(TT measurements don')-.66 E
2.75(to)-.198 G(nly)-2.75 E(come from feedback pack)100.8 202 Q
(ets; the)-.11 E 2.75(yc)-.165 G
(ould also come from other places, such as the SYN)-2.75 E -.165(ex)
100.8 215 S(change.).165 E(Section 4.3, response to feedback pack)100.8
241 Q(ets: Section 4.3 w)-.11 E(as modi\214ed to change the w)-.11 E
(ay that)-.11 E(the recei)100.8 254 Q .33 -.165(ve r)-.275 H
(ate is used in limiting the sender').165 E 2.75(sa)-.605 G(llo)-2.75 E
(wed sending rate, by using the set of)-.275 E(recei)100.8 267 Q .33
-.165(ve r)-.275 H(ate v).165 E(alues of the last tw)-.275 E 2.75(or)
-.11 G(ound-trip times, and initializing the set of recei)-2.75 E .33
-.165(ve r)-.275 H(ate).165 E -.275(va)100.8 280 S(lues by a lar).275 E
(ge v)-.198 E(alue.)-.275 E(The lar)100.8 306 Q
(ger initial sending rate in Section 4.2 is of little use if the recei)
-.198 E -.165(ve)-.275 G 2.75(rs).165 G(ends a feedback)-2.75 E(pack)
100.8 319 Q(et after the \214rst pack)-.11 E(et is recei)-.11 E -.165
(ve)-.275 G(d, and the sender in response reduces the allo).165 E(wed)
-.275 E(sending rate to at most tw)100.8 332 Q 2.75(op)-.11 G(ack)-2.75
E(ets per R)-.11 E(TT)-.66 E 2.75(,w)-.814 G(hich w)-2.75 E
(ould be twice the recei)-.11 E .33 -.165(ve r)-.275 H(ate.).165 E
(Because of the change in the sender')100.8 345 Q 2.75(sp)-.605 G
(rocessing of the recei)-2.75 E .33 -.165(ve r)-.275 H
(ate, the sender no).165 E 2.75(wd)-.275 G(oes)-2.75 E
(not reduce the allo)100.8 358 Q
(wed sending rate to twice the reported recei)-.275 E .33 -.165(ve r)
-.275 H(ate in response to the \214rst).165 E(feedback pack)100.8 371 Q
(et.)-.11 E(The sender ne)100.8 397 Q -.165(ve)-.275 G 2.75(ru).165 G
(ses the recei)-2.75 E .33 -.165(ve r)-.275 H
(ate from a data-limited period to restrict the allo).165 E(wed)-.275 E
(sending rate.)100.8 410 Q(Appendix C discusses this response further)
5.5 E(.)-.605 E(Section 4.4, response to an idle period: F)100.8 436 Q
(ollo)-.165 E(wing Section 5.1 from [RFC4342], this)-.275 E(document sp\
eci\214es that when the sending rate is reduced after an idle period th\
at co)100.8 449 Q -.165(ve)-.165 G(rs the).165 E
(period since the nofeedback timer w)100.8 462 Q(as set, the allo)-.11 E
(wed sending rate is not reduced belo)-.275 E(w)-.275 E
(the initial sending rate.)100.8 475 Q
(Section 4.4, correction from [RFC3448Err].)100.8 501 Q
(RFC 3448 had contradictory te)5.5 E(xt about)-.165 E
(whether the sender halv)100.8 514 Q(ed its sending rate after *tw)-.165
E(o* round-trip times without recei)-.11 E(ving a)-.275 E
(feedback report, or after *four* round-trip times.)100.8 527 Q
(This document clari\214es that the sender)5.5 E(halv)100.8 540 Q
(es its sending rate after four round-trip times without recei)-.165 E
(ving a feedback report)-.275 E([RFC3448Err].)100.8 553 Q
(Section 4.4, clari\214cation for Slo)100.8 579 Q
(w-Start: Section 4.4 w)-.275 E(as clari\214ed to specify that on the)
-.11 E -.165(ex)100.8 592 S(piration of the nofeedback timer).165 E 2.75
(,i)-.44 G 2.75(fp=0)-2.75 G 2.75(,X)-2.75 G(_Bps can')-2.75 E 2.75(tb)
-.198 G 2.75(eu)-2.75 G(sed, because the sender)-2.75 E(doesn')100.8 605
Q 2.75(ty)-.198 G(et ha)-2.75 E .33 -.165(ve a v)-.22 H(alue for X_Bps.)
-.11 E(Section 4.7: credits for unused send time: Section 4.7 has been \
clari\214ed to say that the)100.8 631 Q
(TFRC sender gets to accumulate up to an R)100.8 644 Q
(TT of `credits' for unused send time.)-.66 E(Section)5.5 E(4.7 w)100.8
657 Q(as also re)-.11 E
(written to clarity what is speci\214cation and what is implementation.)
-.275 E(Section 5.4, clari\214cation: Section 5.4 w)100.8 683 Q
(as modi\214ed to clarify the recei)-.11 E -.165(ve)-.275 G(r').165 E
2.75(sc)-.605 G(alculation of)-2.75 E(the a)100.8 696 Q -.165(ve)-.22 G
(rage loss interv).165 E(al when the recei)-.275 E -.165(ve)-.275 G 2.75
(rh).165 G(as not yet seen eight loss interv)-2.75 E(als.)-.275 E
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
197.843(idmer Section)-.44 F 2.75(9. [P)2.75 F(age 29])-.165 E 0 Cg EP
%%Page: 30 30
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E
(Section 5.5, correction: Section 5.5 w)100.8 85 Q
(as corrected to say that the loss interv)-.11 E(al I_0 includes)-.275 E
(all transmitted pack)100.8 98 Q(ets, including lost and mark)-.11 E
(ed pack)-.11 E(ets \(as de\214ned in Section 5.3 in the)-.11 E
(general de\214nition of loss interv)100.8 111 Q(als.\))-.275 E
(Section 5.5, correction from [RFC3448Err]: A line in Section 5.5 w)
100.8 137 Q(as changed from)-.11 E -.814(``)136.8 156 S
(for \(i = 1 to n\) { DF_i = 1; }').814 E(')-.814 E(to)100.8 175 Q -.814
(``)136.8 194 S(for \(i = 0 to n\) { DF_i = 1; }').814 E(')-.814 E
([RFC3448Err].)100.8 213 Q
(Section 5.5, history discounting: THRESHOLD, the lo)100.8 239 Q
(wer bound on the history)-.275 E(discounting parameter DF)100.8 252 Q
2.75(,h)-.88 G(as been changed from 0.5 to 0.25, to allo)-2.75 E 2.75
(wm)-.275 G(ore history)-2.75 E(discounting when the current interv)
100.8 265 Q(al is long.)-.275 E(Section 6, multipe feedback pack)100.8
291 Q(ets: Section 6 no)-.11 E 2.75(wc)-.275 G
(ontains more discussion of procedures)-2.75 E(if the recei)100.8 304 Q
-.165(ve)-.275 G 2.75(rs).165 G(ends multiple feedback pack)-2.75 E
(ets each round-trip time.)-.11 E
(Section 6.3, initialization of the feedback timer: Section 6.3 no)100.8
330 Q 2.75(ws)-.275 G(peci\214es the recei)-2.75 E -.165(ve)-.275 G(r')
.165 E(s)-.605 E
(initialization of the feedback timer if the \214rst data pack)100.8 343
Q(et recei)-.11 E -.165(ve)-.275 G 2.75(dd).165 G(oesn')-2.75 E 2.75(th)
-.198 G -2.475 -.22(av e)-2.75 H(an estimate)2.97 E
(of the round-trip time.)100.8 356 Q
(Section 6.3, a coarse-grained timestamp: Section 6.3 w)100.8 382 Q
(as modi\214ed to incorporate, as an)-.11 E
(option, a coarse-grained timestamp from the sender that increments e)
100.8 395 Q -.165(ve)-.275 G(ry quarter of a).165 E
(round-trip time, instead of a more \214ne-grained timestamp.)100.8 408
Q(This follo)5.5 E(ws RFC 4243.)-.275 E
(Section 6.3.1, after the \214rst loss e)100.8 434 Q -.165(ve)-.275 G
(nt: Section 6.3.1 no).165 E 2.75(ws)-.275 G
(ays that for initializing the loss)-2.75 E
(history after the \214rst loss e)100.8 447 Q -.165(ve)-.275 G
(nt, the recei).165 E -.165(ve)-.275 G 2.75(ru).165 G
(ses the maximum recei)-2.75 E .33 -.165(ve r)-.275 H(ate so f).165 E
(ar)-.11 E 2.75(,i)-.44 G(nstead)-2.75 E(of the recei)100.8 460 Q .33
-.165(ve r)-.275 H(ate in the last round-trip time.).165 E
(Section 6.3.1, if the \214rst data pack)100.8 486 Q
(et is dropped: Section 6.3.1 no)-.11 E 2.75(wc)-.275 G
(ontains a speci\214cation)-2.75 E
(for initializing the loss history if the \214rst data pack)100.8 499 Q
(et sent is lost or ECN-mark)-.11 E(ed.)-.11 E(Section 7, sender)100.8
525 Q(-based v)-.22 E(ariants: Section 7')-.275 E 2.75(sd)-.605 G
(iscussion of sender)-2.75 E(-based v)-.22 E(ariants has been)-.275 E
-.165(ex)100.8 538 S(panded, with reference to RFC 4342.).165 E/F1 11
/Times-Bold@0 SF(10.)72 577 Q/F2 14/Times-Bold@0 SF
(Security Considerations)5.5 E F0
(TFRC is not a transport protocol in its o)100.8 593.6 Q(wn right, b)
-.275 E(ut a congestion control mechanism that)-.22 E
(is intended to be used in conjunction with a transport protocol.)100.8
606.6 Q(Therefore security)5.5 E
(primarily needs to be considered in the conte)100.8 619.6 Q
(xt of a speci\214c transport protocol and its)-.165 E
(authentication mechanisms.)100.8 632.6 Q
(Congestion control mechanisms can potentially be e)100.8 649.2 Q
(xploited to create denial of service.)-.165 E
(This may occur through spoofed feedback.)100.8 662.2 Q(Thus an)5.5 E
2.75(yt)-.165 G(ransport protocol that uses TFRC)-2.75 E(should tak)
100.8 675.2 Q 2.75(ec)-.11 G
(are to ensure that feedback is only accepted from the recei)-2.75 E
-.165(ve)-.275 G 2.75(ro).165 G 2.75(ft)-2.75 G(he data.)-2.75 E(The)5.5
E(precise mechanism to achie)100.8 688.2 Q .33 -.165(ve t)-.275 H
(his will ho).165 E(we)-.275 E -.165(ve)-.275 G 2.75(rd).165 G
(epend on the transport protocol itself.)-2.75 E(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 192.343(idmer Section)-.44
F 2.75(10. [P)2.75 F(age 30])-.165 E 0 Cg EP
%%Page: 31 31
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(In addition, congestion contro\
l mechanisms may potentially be manipulated by a greedy)100.8 85 Q
(recei)100.8 98 Q -.165(ve)-.275 G 2.75(rt).165 G(hat wishes to recei)
-2.75 E .33 -.165(ve m)-.275 H(ore than its f).165 E(air share of netw)
-.11 E(ork bandwidth.)-.11 E 2.75(Ar)5.5 G(ecei)-2.75 E -.165(ve)-.275 G
(r).165 E(might do this by claiming to ha)100.8 111 Q .33 -.165(ve r)
-.22 H(ecei).165 E -.165(ve)-.275 G 2.75(dp).165 G(ack)-2.75 E
(ets that in f)-.11 E(act were lost due to congestion.)-.11 E
(Possible defenses ag)100.8 124 Q(ainst such a recei)-.055 E -.165(ve)
-.275 G 2.75(rw).165 G(ould normally include some form of nonce that)
-2.86 E(the recei)100.8 137 Q -.165(ve)-.275 G 2.75(rm).165 G
(ust feed back to the sender to pro)-2.75 E .33 -.165(ve r)-.165 H 2.75
(eceipt. Ho).165 F(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H
(he details of such a).44 E(nonce w)100.8 150 Q(ould depend on the tran\
sport protocol, and in particular on whether the transport)-.11 E
(protocol is reliable or unreliable.)100.8 163 Q 1.76 -.88(We e)100.8
179.6 T(xpect that protocols incorporating ECN with TFRC will also w)
.715 E(ant to incorporate)-.11 E(feedback from the recei)100.8 192.6 Q
-.165(ve)-.275 G 2.75(rt).165 G 2.75(ot)-2.75 G
(he sender using the ECN nonce [RFC3540].)-2.75 E(The ECN)8.25 E(nonce \
is a modi\214cation to ECN that protects the sender from the accidental\
 or malicious)100.8 205.6 Q(concealment of mark)100.8 218.6 Q(ed pack)
-.11 E 2.75(ets. Ag)-.11 F(ain, the details of such a nonce w)-.055 E
(ould depend on the)-.11 E
(transport protocol, and are not addressed in this document.)100.8 231.6
Q/F1 11/Times-Bold@0 SF(11.)72 270.6 Q/F2 14/Times-Bold@0 SF(IAN)5.5 E
3.5(AC)-.28 G(onsiderations)-3.5 E F0(There are no IAN)100.8 287.2 Q
2.75(Aa)-.385 G(ctions required for this document.)-2.75 E F1(12.)72
326.2 Q F2(Ackno)5.5 E(wledgments)-.14 E F0 1.76 -.88(We w)100.8 342.8 T
(ould lik).77 E 2.75(et)-.11 G 2.75(oa)-2.75 G(ckno)-2.75 E
(wledge feedback and discussions on equation-based congestion)-.275 E(c\
ontrol with a wide range of people, including members of the Reliable M\
ulticast Research)100.8 355.8 Q(Group, the Reliable Multicast T)100.8
368.8 Q(ransport W)-.385 E(orking Group, and the End-to-End Research)
-.88 E 5.5(Group. W)100.8 381.8 R 2.75(ew)-.88 G(ould lik)-2.86 E 2.75
(et)-.11 G 2.75(ot)-2.75 G(hank Dado Colussi, Gorry F)-2.75 E
(airhurst, Ladan Gharai, W)-.165 E(im)-.44 E(Heirman, Eddie K)100.8
394.8 Q(ohler)-.385 E 2.75(,K)-.44 G(en Lofgren, Mik)-3.025 E 2.75(eL)
-.11 G(uby)-2.75 E 2.75(,I)-.715 G(an McDonald, Michele R., Gerrit)-2.75
E(Renk)100.8 407.8 Q(er)-.11 E 2.75(,A)-.44 G
(rjuna Sathiaseelan, Vladica Stanisic, Randall Ste)-2.75 E -.11(wa)-.275
G(rt, Eduardo Urzaiz, Shushan).11 E -.88(We)100.8 420.8 S(n, and W).88 E
(endy Lee \(lhh@zsu.edu.cn\) for feedback on earlier v)-.88 E
(ersions of this document,)-.165 E(and to thank Mark Allman for his e)
100.8 433.8 Q(xtensi)-.165 E .33 -.165(ve f)-.275 H
(eedback from using [RFC3448] to produce a).165 E -.11(wo)100.8 446.8 S
(rking implementation.).11 E F1(A.)72 472.8 Q F2 -1.288(Te)5.5 G
(rminology)1.288 E F0(This document uses the follo)100.8 489.4 Q
(wing terms.)-.275 E 8.934(DF: Discount)100.8 506 R -.11(fa)2.75 G
(ctor for a loss interv).11 E(al \(Section 5.5\).)-.275 E(initial_rate:)
100.8 522.6 Q(Allo)129.6 535.6 Q(wed initial sending rate.)-.275 E
(last_counter:)100.8 552.2 Q(Greatest recei)129.6 565.2 Q -.165(ve)-.275
G 2.75(dv).165 G(alue of the windo)-3.025 E 2.75(wc)-.275 G
(ounter \(Section 6.3\).)-2.75 E(min_rate:)100.8 581.8 Q
(Minimum transmit rate \(Section 4.3\).)129.6 594.8 Q 17.492(n: Number)
100.8 611.4 R(of loss interv)2.75 E(als.)-.275 E(NDUP)100.8 628 Q -.44
(AC)-1.012 G(K:).44 E
(Number of dupacks for inferring loss \(constant\) \(Section 5.1\).)
129.6 641 Q(nofeedback timer:)100.8 657.6 Q(Sender)129.6 670.6 Q
(-side timer \(Section 4\).)-.22 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)
-.11 E(adh)-.165 E(ye/W)-.055 E 195.401(idmer Section)-.44 F 2.75(A. [P)
2.75 F(age 31])-.165 E 0 Cg EP
%%Page: 32 32
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 17.492(p: Estimated)100.8 85 R
(Loss Ev)2.75 E(ent Rate.)-.165 E(p_pre)100.8 101.6 Q(v:)-.275 E(Pre)
129.6 114.6 Q(vious v)-.275 E(alue of p \(Section 6.1\).)-.275 E 17.492
(q: Filter)100.8 131.2 R(constant for R)2.75 E
(TT \(constant\) \(Section 4.3\).)-.66 E 11.992(q2: Filter)100.8 147.8 R
(constant for long-term R)2.75 E(TT \(constant\) \(Section 4.6\).)-.66 E
15.655(R: Estimated)100.8 164.4 R(path round-trip time.)2.75 E
(R_sample:)100.8 181 Q(Measured path R)129.6 194 Q(TT \(Section 4.3\).)
-.66 E(R_sqmean:)100.8 210.6 Q
(Long-term estimate of the square root of the R)129.6 223.6 Q
(TT \(Section 4.6\).)-.66 E(reco)100.8 240.2 Q -.165(ve)-.165 G(r_rate:)
.165 E(Allo)129.6 253.2 Q(wed rate for resuming after an idle period.)
-.275 E(recv_limit;)100.8 269.8 Q
(Limit on sending rate computed from the recei)129.6 282.8 Q .33 -.165
(ve r)-.275 H(ate.).165 E 18.713(s: Nominal)100.8 299.4 R(pack)2.75 E
(et size in bytes.)-.11 E 16.876(S: Sequence)100.8 316 R(number)2.75 E
(.)-.605 E(t_delay:)100.8 332.6 Q
(Reported time delay between receipt of the last pack)129.6 345.6 Q
(et at the recei)-.11 E -.165(ve)-.275 G 2.75(ra).165 G(nd the)-2.75 E
(generation of the feedback pack)129.6 358.6 Q(et \(Section 3.2.2\).)
-.11 E(t_delta:)100.8 375.2 Q -.165(Pa)129.6 388.2 S(rameter for \215e)
.165 E(xibility in send time \(Section 4.7\).)-.165 E(t_gran:)100.8
404.8 Q(Schedular granularity \(constant\) \(Section 4.7\).)129.6 417.8
Q 2.818(t_ipi: Inter)100.8 434.4 R(-pack)-.22 E(et interv)-.11 E
(al for sending pack)-.275 E(ets \(Section 4.7\).)-.11 E(t_mbi:)100.8
451 Q(Maximum R)129.6 464 Q .396 -.198(TO v)-.66 H
(alue of TCP \(constant\) \(Section 4.3\).)-.077 E 11.376(tld: T)100.8
480.6 R(ime Last Doubled \(Section 4.2\).)-.385 E(t_no)100.8 497.2 Q(w:)
-.275 E(Current time \(Section 4.3\).)129.6 510.2 Q(t_R)100.8 526.8 Q
-.198(TO)-.66 G(:).198 E(Estimated R)129.6 539.8 Q .396 -.198(TO o)-.66
H 2.75(fT).198 G(CP \(Section 4.3\).)-2.75 E 15.05(X: Allo)100.8 556.4 R
(wed transmit rate, as limited by the recei)-.275 E .33 -.165(ve r)-.275
H(ate.).165 E(X_Bps:)100.8 573 Q
(Calculated sending rate in bytes per second \(Section 3.1\).)129.6 586
Q(X_pps:)100.8 602.6 Q(Calculated sending rate in pack)129.6 615.6 Q
(ets per second \(Section 3.1\).)-.11 E(X_recv:)100.8 632.2 Q
(Estimated recei)129.6 645.2 Q .33 -.165(ve r)-.275 H(ate at the recei)
.165 E -.165(ve)-.275 G 2.75(r\().165 G(Section 3.2.2\).)-2.75 E
(X_inst:)100.8 661.8 Q(Instantaneous allo)129.6 674.8 Q
(wed transmit rate \(Section 4.6\).)-.275 E(Handle)72 733 Q(y/Flo)-.165
E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 195.401(idmer Section)-.44 F 2.75
(A. [P)2.75 F(age 32])-.165 E 0 Cg EP
%%Page: 33 33
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(W_init:)100.8 85 Q
(TCP initial windo)129.6 98 Q 2.75(w\()-.275 G
(constant\) \(Section 4.2\).)-2.75 E/F1 11/Times-Bold@0 SF(B.)72 124 Q
/F2 14/Times-Bold@0 SF(The Initial V)5.5 E(alue of the Nofeedback T)
-1.288 E(imer)-.252 E F0(Wh)100.8 150 Q 2.75(yi)-.055 G 2.75(st)-2.75 G
(he initial v)-2.75 E(alue of TFRC')-.275 E 2.75(sn)-.605 G
(ofeedback timer set to tw)-2.75 E 2.75(os)-.11 G
(econds, instead of the)-2.75 E(recommended initial v)100.8 163 Q
(alue of three seconds for TCP')-.275 E 2.75(sr)-.605 G(etransmit timer)
-2.75 E 2.75(,f)-.44 G(rom [RFC2988]?)-2.75 E(There isn')100.8 176 Q
2.75(ta)-.198 G .33 -.165(ny p)-2.75 H(articular reason wh).165 E 2.75
(yT)-.055 G(FRC')-2.75 E 2.75(sn)-.605 G(ofeedback timer should ha)-2.75
E .33 -.165(ve t)-.22 H(he same).165 E(initial v)100.8 189 Q
(alue as TCP')-.275 E 2.75(sr)-.605 G(etransmit timer)-2.75 E 5.5(.T)
-.605 G(CP')-5.5 E 2.75(sr)-.605 G
(etransmit timer is used not only to reduce)-2.75 E
(the sending rate in response to congestion, b)100.8 202 Q
(ut also to retransmit a pack)-.22 E(et that is assumed to)-.11 E(ha)
100.8 215 Q .33 -.165(ve b)-.22 H(een dropped in the netw).165 E 2.75
(ork. In)-.11 F(contrast, TFRC')2.75 E 2.75(sn)-.605 G
(ofeedback timer is only used to)-2.75 E(reduce the allo)100.8 228 Q
(wed sending rate, not to trigger the sending of a ne)-.275 E 2.75(wp)
-.275 G(ack)-2.75 E 2.75(et. As)-.11 F 2.75(ar)2.75 G(esult,)-2.75 E
(there is no danger to the netw)100.8 241 Q(ork for the initial v)-.11 E
(alue of TFRC')-.275 E 2.75(sn)-.605 G(ofeedback timer to be)-2.75 E
(smaller than the recommended initial v)100.8 254 Q(alue for TCP')-.275
E 2.75(sr)-.605 G(etransmit timer)-2.75 E(.)-.605 E(Further)100.8 280 Q
2.75(,w)-.44 G(hen the nofeedback timer has not yet e)-2.75 E
(xpired, TFRC has a more slo)-.165 E(wly-)-.275 E
(responding congestion control mechanism that TCP)100.8 293 Q 2.75(,a)
-1.221 G(nd TFRC')-2.75 E 2.75(su)-.605 G(se of the recei)-2.75 E .33
-.165(ve r)-.275 H(ate for).165 E(limiting the sending rate is some)
100.8 306 Q(what less precise than TCP')-.275 E 2.75(su)-.605 G
(se of windo)-2.75 E(ws and ack-)-.275 E(clocking, so the nofeedback ti\
mer is a particularly important safety mechanism for TFRC.)100.8 319 Q
-.165(Fo)100.8 332 S 2.75(ra).165 G
(ll of these reasons, it is perfectly reasonable for TFRC')-2.75 E 2.75
(sn)-.605 G(ofeedback timer to ha)-2.75 E .33 -.165(ve a)-.22 H
(smaller initial v)100.8 345 Q(alue than that of TCP')-.275 E 2.75(sr)
-.605 G(etransmit timer)-2.75 E(.)-.605 E F1(C.)72 371 Q F2
(Response to Idle or Data-limited P)5.5 E(eriods)-.28 E F0(Future w)
100.8 397 Q(ork could e)-.11 E
(xplore alternate responses to using the recei)-.165 E .33 -.165(ve r)
-.275 H(ate during a data-).165 E(limited period.)100.8 410 Q F1(C.1.)72
436 Q/F3 13/Times-Bold@0 SF(Long Idle or Data-limited P)5.5 E(eriods)
-.26 E F0 -.88(Ta)100.8 462 S(ble 1 summarizes the response of Standard\
 TCP [RFC2581], TCP with Congestion).88 E -.44(Wi)100.8 475 S(ndo).44 E
2.75(wV)-.275 G(alidation [RFC2861], Standard TFRC [RFC3448], and Re)
-3.971 E(vised TFRC \(this)-.275 E
(document\) in response to long idle or data-limited periods.)100.8 488
Q -.165(Fo)5.5 G 2.75(rt).165 G(he purposes of this)-2.75 E
(section, we de\214ne a long period as a period of at least an R)100.8
501 Q -.198(TO)-.66 G(.).198 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E
(adh)-.165 E(ye/W)-.055 E 187.756(idmer Section)-.44 F 2.75(C.1. [P)2.75
F(age 33])-.165 E 0 Cg EP
%%Page: 34 34
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 22(Protocol Long)106.3 85 R
(idle periods)2.75 E(Long data-limited periods)16.5 E 5.5
(-------------- --------------------)100.8 98 R(----------------------)
13.75 E(Standard TCP:)100.8 111 Q -.44(Wi)19.25 G(ndo).44 E 2.75(w-)
-.275 G 2.75(>i)-2.75 G 11(nitial. No)-2.75 F(change in windo)2.75 E
-.715(w.)-.275 G(TCP with CWV)100.8 137 Q 24.75(:H)-.814 G(alv)-24.75 E
2.75(ew)-.165 G(indo)-2.75 E 24.75(wR)-.275 G(educe windo)-24.75 E 2.75
(wh)-.275 G(alf w)-2.75 E(ay)-.11 E(\(not belo)144.8 150 Q 2.75(wi)-.275
G(nitial cwnd\).)-2.75 E(to used windo)13.75 E -.715(w.)-.275 G
(Standard TFRC:)100.8 176 Q(Halv)22 E 2.75(er)-.165 G 35.75(ate Rate)
-2.75 F(limited to)2.75 E(\(not belo)147.55 189 Q 2.75(w1p)-.275 G
(kt/64 sec\).)-2.75 E(twice recei)8.25 E .33 -.165(ve r)-.275 H(ate.)
.165 E(One R)147.55 202 Q(TT after sending pkt,)-.66 E
(rate is limited by X_recv)147.55 215 Q(.)-.715 E(Re)100.8 241 Q
(vised TFRC:)-.275 E(Halv)24.75 E 2.75(er)-.165 G 33(ate Rate)-2.75 F
(not limited to)2.75 E(\(not belo)147.55 254 Q 2.75(wi)-.275 G
(nitial rate\).)-2.75 E(twice recei)8.25 E .33 -.165(ve r)-.275 H(ate.)
.165 E -.88(Ta)106.3 273 S
(ble 1: Response to long idle or data-limited periods.).88 E
(Standard TCP after long idle periods: F)100.8 299 Q(or Standard TCP)
-.165 E 2.75(,[)-1.221 G(RFC2581] speci\214es that TCP)-2.75 E
(SHOULD set the congestion windo)100.8 312 Q 2.75(wt)-.275 G 2.75(on)
-2.75 G 2.75(om)-2.75 G(ore than the initial windo)-2.75 E 2.75(wa)-.275
G(fter an idle)-2.75 E(period of at least an R)100.8 325 Q -.198(TO)-.66
G(.).198 E(Standard TCP after long data-limited periods: Standard TCP [\
RFC2581] does not reduce)100.8 351 Q(TCP')100.8 364 Q 2.75(sc)-.605 G
(ongestion windo)-2.75 E 2.75(wa)-.275 G
(fter a data-limited period, when the congestion windo)-2.75 E 2.75(wi)
-.275 G 2.75(sn)-2.75 G(ot)-2.75 E(fully used.)100.8 377 Q(Standard TCP\
 in [RFC2581] uses the FlightSize, the amount of outstanding)5.5 E
(data in the netw)100.8 390 Q(ork, only in setting the slo)-.11 E
(w-start threshold after a retransmit timeout.)-.275 E
(Standard TCP is not limited by TCP')100.8 403 Q 2.75(sa)-.605 G
(ck-clocking mechanism during a data-limited)-2.75 E(period.)100.8 416 Q
(Standard TCP')100.8 442 Q 2.75(sl)-.605 G
(ax response to a data-limited period is quite dif)-2.75 E
(ferent from its stringent)-.275 E(response to an idle period.)100.8 455
Q(TCP with Congestion W)100.8 481 Q(indo)-.44 E 2.75(wV)-.275 G
(alidation \(CWV\) after long idle periods: As an)-3.971 E -.165(ex)
100.8 494 S(perimental alternati).165 E -.165(ve)-.275 G 2.75(,[).165 G
(RFC2861] speci\214es a more moderate response to an idle period)-2.75 E
(than that of Standard TCP)100.8 507 Q 2.75(,w)-1.221 G
(here during an idle period the TCP sender halv)-2.75 E(es cwnd after)
-.165 E(each R)100.8 520 Q -.198(TO)-.66 G 2.75(,d).198 G -.275(ow)-2.75
G 2.75(nt).275 G 2.75(ot)-2.75 G(he initial cwnd.)-2.75 E
(TCP with Congestion W)100.8 546 Q(indo)-.44 E 2.75(wV)-.275 G
(alidation after long data-limited periods: As an)-3.971 E -.165(ex)
100.8 559 S(perimental alternati).165 E -.165(ve)-.275 G 2.75(,[).165 G
(RFC2861] speci\214es a more stringent response to a data-limited)-2.75
E(period than that of Standard TCP)100.8 572 Q 2.75(,w)-1.221 G
(here after each R)-2.75 E .396 -.198(TO s)-.66 H
(econds of a data-limited period,).198 E(the congestion windo)100.8 585
Q 2.75(wi)-.275 G 2.75(sr)-2.75 G(educed half w)-2.75 E(ay do)-.11 E
(wn to the windo)-.275 E 2.75(wt)-.275 G(hat is actually used.)-2.75 E(\
The response of TCP with CWV to an idle period is similar to its respon\
se to a data-limited)100.8 611 Q 2.75(period. TCP)100.8 624 R
(with CWV is less restricti)2.75 E .33 -.165(ve t)-.275 H
(han Standard TCP in response to an idle period,).165 E
(and more restricti)100.8 637 Q .33 -.165(ve t)-.275 H
(han Standard TCP in response to a data-limited period.).165 E
(Standard TFRC after long idle periods: F)100.8 663 Q
(or Standard TFRC, [RFC3448] speci\214es that the)-.165 E(allo)100.8 676
Q(wed sending rate is halv)-.275 E(ed after each R)-.165 E .396 -.198
(TO s)-.66 H(econds of an idle period.).198 E(The allo)5.5 E(wed)-.275 E
(sending rate is not reduced belo)100.8 689 Q 2.75(wo)-.275 G(ne pack)
-2.75 E(et in 64 seconds.)-.11 E(After an idle period, the \214rst)5.5 E
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
187.756(idmer Section)-.44 F 2.75(C.1. [P)2.75 F(age 34])-.165 E 0 Cg EP
%%Page: 35 35
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(feedback pack)100.8 85 Q
(et recei)-.11 E -.165(ve)-.275 G 2.75(sr).165 G(eports a recei)-2.75 E
.33 -.165(ve r)-.275 H(ate of one pack).165 E
(et per round-trip time, and this)-.11 E(recei)100.8 98 Q .33 -.165
(ve r)-.275 H(ate is used to limit the sending rate.).165 E
(Standard TFRC ef)5.5 E(fecti)-.275 E -.165(ve)-.275 G(ly slo).165 E
(w-starts up)-.275 E(from this allo)100.8 111 Q(wed sending rate.)-.275
E(Standard TFRC after long data-limited periods: [RFC3448] does not dis\
tinguish between)100.8 137 Q(data-limited and not-data-limited periods.)
100.8 150 Q(As a consequence, the allo)5.5 E(wed sending rate is)-.275 E
(limited to at most twice the recei)100.8 163 Q .33 -.165(ve r)-.275 H
(ate during and after a data-limited period.).165 E(This is a)5.5 E
-.165(ve)100.8 176 S(ry restricti).165 E .33 -.165(ve r)-.275 H
(esponse, more restricti).165 E .33 -.165(ve t)-.275 H
(han that of either Standard TCP or of TCP with).165 E(CWV)100.8 189 Q
(.)-1.419 E(Re)100.8 215 Q(vised TFRC after long idle periods: F)-.275 E
(or Re)-.165 E(vised TFRC, this document speci\214es that the)-.275 E
(allo)100.8 228 Q(wed sending rate is halv)-.275 E(ed after each R)-.165
E .396 -.198(TO s)-.66 H(econds of an idle period.).198 E(The allo)5.5 E
(wed)-.275 E(sending rate is not reduced belo)100.8 241 Q 2.75(wt)-.275
G(he initial sending rate as the result of an idle period.)-2.75 E
(The \214rst feedback pack)100.8 254 Q(et recei)-.11 E -.165(ve)-.275 G
2.75(da).165 G(fter the idle period reports a recei)-2.75 E .33 -.165
(ve r)-.275 H(ate of one pack).165 E(et)-.11 E(per round-trip time.)
100.8 267 Q(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H
(he Re).44 E(vised TFRC sender does not use this recei)-.275 E .33 -.165
(ve r)-.275 H(ate for).165 E(limiting the sending rate.)100.8 280 Q
(Thus, Re)5.5 E(vised TFRC dif)-.275 E
(fers from Standard TFRC in the lo)-.275 E(wer)-.275 E(limit used in th\
e reduction of the sending rate, and in the better response to the \214\
rst)100.8 293 Q(feedback pack)100.8 306 Q(et recei)-.11 E -.165(ve)-.275
G 2.75(da).165 G(fter the idle period.)-2.75 E(Re)100.8 332 Q
(vised TFRC after long data-limited periods: F)-.275 E(or Re)-.165 E
(vised TFRC, this document)-.275 E
(distinguishs between data-limited and not-data-limited periods.)100.8
345 Q(As speci\214ed in Section)5.5 E(4.3, Re)100.8 358 Q
(vised TFRC does not reduce the allo)-.275 E
(wed sending rate in response to the recei)-.275 E .33 -.165(ve r)-.275
H(ate).165 E(during a data-limited period.)100.8 371 Q
(This is perhaps an o)5.5 E -.165(ve)-.165 G(rly-lax response, b).165 E
(ut it is similar to the)-.22 E(response of Standard TCP)100.8 384 Q
2.75(,a)-1.221 G(nd is quite dif)-2.75 E(ferent from the v)-.275 E
(ery restricti)-.165 E .33 -.165(ve r)-.275 H(esponse of).165 E
(Standard TFRC to a data-limited period.)100.8 397 Q(Reco)100.8 423 Q
-.165(ve)-.165 G(ry after idle or data-limited periods: When TCP reduce\
s the congestion windo).165 E(w)-.275 E
(after an idle or data-utilized period, TCP can set the slo)100.8 436 Q
(w-start threshold ssthresh to allo)-.275 E(w)-.275 E
(the TCP sender to slo)100.8 449 Q(w-start back up to)-.275 E -.11(wa)
-.275 G(rds its old sending rate when the idle or data-).11 E
(limited period is o)100.8 462 Q -.165(ve)-.165 G 3.96 -.605(r. H).165 H
-.275(ow).605 G -2.365 -.275(ev e).275 H 2.75(ri).275 G 2.75(nT)-2.75 G
(FRC, e)-2.75 E -.165(ve)-.275 G 2.75(nw).165 G(hen the TFRC sender')
-2.75 E 2.75(ss)-.605 G(ending rate is)-2.75 E
(restricted by twice the pre)100.8 475 Q(vious recei)-.275 E .33 -.165
(ve r)-.275 H(ate, this results in the sender being able to double).165
E(the sending rate from one round-trip time to the ne)100.8 488 Q
(xt, if permitted by the throughput)-.165 E 2.75(equation. Thus,)100.8
501 R(TFRC doesn')2.75 E 2.75(tn)-.198 G(eed a mechanism such as TCP')
-2.75 E 2.75(ss)-.605 G(etting of ssthresh to)-2.75 E(allo)100.8 514 Q
2.75(was)-.275 G(lo)-2.75 E
(w-start after an idle or data-limited period.)-.275 E -.165(Fo)100.8
540 S 2.75(rf).165 G(uture w)-2.75 E(ork, one a)-.11 E -.165(ve)-.22 G
(nue to e).165 E(xplore w)-.165 E(ould be the addition of Congestion W)
-.11 E(indo)-.44 E(w)-.275 E -1.221(Va)100.8 553 S
(lidation mechanisms for TFRC')1.221 E 2.75(sr)-.605 G
(esponse to data-limited periods.)-2.75 E(Currently)5.5 E 2.75(,f)-.715
G(ollo)-2.75 E(wing)-.275 E(Standard TCP)100.8 566 Q 2.75(,d)-1.221 G
(uring data-limited periods Re)-2.75 E
(vised TFRC does not limit its allo)-.275 E(wed)-.275 E
(sending rate as a function of the recei)100.8 579 Q .33 -.165(ve r)
-.275 H(ate.).165 E/F1 11/Times-Bold@0 SF(C.2.)72 605 Q/F2 13
/Times-Bold@0 SF(Short Idle or Data-limited P)5.5 E(eriods)-.26 E F0
-.88(Ta)100.8 631 S(ble 2 summarizes the response of Standard TCP [RFC2\
581], TCP with Congestion).88 E -.44(Wi)100.8 644 S(ndo).44 E 2.75(wV)
-.275 G(alidation [RFC2861], Standard TFRC [RFC3448], and Re)-3.971 E
(vised TFRC \(this)-.275 E
(document\) in response to short idle or data-limited periods.)100.8 657
Q -.165(Fo)5.5 G 2.75(rt).165 G(he purposes of this)-2.75 E
(section, we de\214ne a short period as a period of less than an R)100.8
670 Q(TT)-.66 E(.)-.814 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)
-.165 E(ye/W)-.055 E 187.756(idmer Section)-.44 F 2.75(C.2. [P)2.75 F
(age 35])-.165 E 0 Cg EP
%%Page: 36 36
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 22(Protocol Short)106.3 85 R
(idle periods)2.75 E(Short data-limited periods)8.25 E 5.5
(-------------- --------------------)100.8 98 R(----------------------)
13.75 E(Standard TCP:)100.8 111 Q(Send a b)11 E(urst up to cwnd.)-.22 E
(Send a b)5.5 E(urst up to cwnd.)-.22 E(TCP with CWV)100.8 137 Q 11(:S)
-.814 G(end a b)-11 E(urst up to cwnd.)-.22 E(Send a b)5.5 E
(urst up to cwnd.)-.22 E(Standard TFRC:)100.8 163 Q 68.75(??)35.75 G(Re)
100.8 189 Q(vised TFRC:)-.275 E(Send a b)24.75 E 38.5(urst Send)-.22 F
2.75(ab)2.75 G(urst)-2.97 E(\(up to an R)158.55 202 Q(TT of)-.66 E
(\(up to an R)30.25 E(TT of)-.66 E(unused send credits\).)153.05 215 Q
(unused send credits\).)16.5 E -.88(Ta)106.3 234 S
(ble 2: Response to short idle or data-limited periods.).88 E -.88(Ta)
100.8 260 S(ble 2 sho).88 E(ws that Re)-.275 E
(vised TFRC has a similar response to that of Standard TCP and of)-.275
E(TCP with CWV to a short idle or data-limited period.)100.8 273 Q -.165
(Fo)5.5 G 2.75(ras).165 G(hort idle or data-limited)-2.75 E
(period, TCP is limited only by the size of the unused congestion windo)
100.8 286 Q 1.43 -.715(w, a)-.275 H(nd Re).715 E(vised)-.275 E
(TFRC is limited only by the number of unused send credits \(up to an R)
100.8 299 Q(TT')-.66 E 2.75(sw)-.605 G 2.75(orth\). F)-2.86 F(or)-.165 E
(Standard TFRC, [RFC3448] did not e)100.8 312 Q
(xplicitly specify the beha)-.165 E(vior with respect to unused)-.22 E
(send credits.)100.8 325 Q/F1 11/Times-Bold@0 SF(C.3.)72 351 Q/F2 13
/Times-Bold@0 SF(Moderate Idle or Data-limited P)5.5 E(eriods)-.26 E F0
-.88(Ta)100.8 377 S(ble 3 summarizes the response of Standard TCP [RFC2\
581], TCP with Congestion).88 E -.44(Wi)100.8 390 S(ndo).44 E 2.75(wV)
-.275 G(alidation [RFC2861], Standard TFRC [RFC3448], and Re)-3.971 E
(vised TFRC \(this)-.275 E
(document\) in response to moderate idle or data-limited periods.)100.8
403 Q -.165(Fo)5.5 G 2.75(rt).165 G(he purposes of this)-2.75 E
(section, we de\214ne a moderate period as a period greater than an R)
100.8 416 Q(TT)-.66 E 2.75(,b)-.814 G(ut less than an R)-2.97 E -.198
(TO)-.66 G(.).198 E 13.75(Protocol Moderate)106.3 448 R(idle periods)
2.75 E(Moderate data-limited periods)5.5 E 5.5
(------------- ---------------------)100.8 461 R
(-------------------------)16.5 E(Standard TCP:)100.8 474 Q(Send a b)11
E(urst up to cwnd.)-.22 E(Send a b)5.5 E(urst up to cwnd.)-.22 E
(TCP with CWV)100.8 500 Q 11(:S)-.814 G(end a b)-11 E(urst up to cwnd.)
-.22 E(Send a b)5.5 E(urst up to cwnd.)-.22 E(Standard TFRC:)100.8 526 Q
52.25(?L)35.75 G(imited by X_recv)-52.25 E(.)-.715 E(Re)100.8 552 Q
(vised TFRC:)-.275 E(Send a b)24.75 E 38.5(urst Send)-.22 F 2.75(ab)2.75
G(urst)-2.97 E(\(up to an R)158.55 565 Q(TT of)-.66 E(\(up to an R)30.25
E(TT of)-.66 E(unused send credits\).)153.05 578 Q
(unused send credits\).)16.5 E -.88(Ta)106.3 597 S
(ble 3: Response to moderate idle or data-limited periods.).88 E -.88
(Ta)100.8 623 S(ble 3 sho).88 E(ws that Re)-.275 E
(vised TFRC has a similar response to that of Standard TCP and of)-.275
E(TCP with CWV to a moderate idle or data-limited period.)100.8 636 Q
-.165(Fo)5.5 G 2.75(ram).165 G(oderate idle or data-)-2.75 E(limited pe\
riod, TCP is limited only by the size of the unused congestion windo)
100.8 649 Q 4.18 -.715(w. F)-.275 H(or a).55 E(moderate idle period, Re)
100.8 662 Q
(vised TFRC is limited only by the number of unused send credits)-.275 E
(\(up to an R)100.8 675 Q(TT')-.66 E 2.75(sw)-.605 G 2.75(orth\). F)
-2.86 F(or a moderate data-limited period, Standard TCP w)-.165 E
(ould be)-.11 E(limited by X_recv from the most recent feedback pack)
100.8 688 Q 2.75(et. In)-.11 F(contrast, Re)2.75 E(vised TFRC isn')-.275
E(t)-.198 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)
-.055 E 187.756(idmer Section)-.44 F 2.75(C.3. [P)2.75 F(age 36])-.165 E
0 Cg EP
%%Page: 37 37
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(limited by the recei)100.8 85 Q
.33 -.165(ve r)-.275 H(ate from data-limited periods that co).165 E
-.165(ve)-.165 G 2.75(ra).165 G 2.75(ne)-2.75 G
(ntire feedback period of)-2.75 E 2.75(ar)100.8 98 S(ound-trip time.)
-2.75 E -.165(Fo)5.5 G 2.75(rS).165 G(tandard TFRC, [RFC3448] did not e)
-2.75 E(xplicitly specify the beha)-.165 E(vior)-.22 E
(with respect to unused send credits.)100.8 111 Q/F1 11/Times-Bold@0 SF
(C.4.)72 137 Q/F2 13/Times-Bold@0 SF(Other P)5.5 E(atter)-.13 E(ns)-.195
E F0(Other possible patterns to consider in e)100.8 163 Q -.275(va)-.275
G(luting Re).275 E(vised TFRC w)-.275 E(ould be to compare the)-.11 E
(beha)100.8 176 Q(vior of TCP)-.22 E 2.75(,S)-1.221 G
(tandard TFRC, and Re)-2.75 E
(vised TFRC for connections with alternating b)-.275 E(usy)-.22 E(and i\
dle periods, alternating idle and data-limited periods, or with idle or\
 data-limited)100.8 189 Q(periods during Slo)100.8 202 Q(w-Start,)-.275
E/F3 14/Times-Bold@0 SF(Normati)72 228 Q .28 -.14(ve R)-.14 H(efer).14 E
(ences)-.252 E F0 29.634([RFC3448] M.)113.1 244.6 R(Handle)2.75 E 1.43
-.715(y, S)-.165 H 2.75(.F).715 G(lo)-2.75 E(yd, J. P)-.11 E(adh)-.165 E
(ye, and J. W)-.055 E(idmer)-.44 E 2.75(,T)-.44 G(CP Friendly Rate)-2.75
E(Control \(TFRC\): Protocol Speci\214cation, RFC 3448, January 2003.)
195.6 257.6 Q F3(Inf)72 283.6 Q(ormational Refer)-.35 E(ences)-.252 E F0
40.634([BRS99] Balakrishnan,)113.1 300.2 R
(H., Rahul, H., and Seshan, S., "An Inte)2.75 E(grated)-.165 E
(Congestion Management Architecture for Internet Hosts," Proc.)195.6
313.2 Q -.44(AC)195.6 326.2 S 2.75(MS).44 G
(IGCOMM, Cambridge, MA, September 1999.)-2.75 E 30.866([FHPW00] S.)113.1
342.8 R(Flo)2.75 E(yd, M. Handle)-.11 E 1.43 -.715(y, J)-.165 H 2.75(.P)
.715 G(adh)-2.915 E(ye, and J. W)-.055 E(idmer)-.44 E 2.75(,")-.44 G
(Equation-Based)-2.75 E
(Congestion Control for Unicast Applications", August 2000, Proc)195.6
355.8 Q(SIGCOMM 2000.)195.6 368.8 Q 25.982([FHPW00a] S.)113.1 385.4 R
(Flo)2.75 E(yd, M. Handle)-.11 E 1.43 -.715(y, J)-.165 H 2.75(.P).715 G
(adh)-2.915 E(ye, and J. W)-.055 E(idmer)-.44 E 2.75(,")-.44 G
(Equation-Based)-2.75 E
(Congestion Control for Unicast Applications: the Extended V)195.6 398.4
Q(ersion",)-1.221 E(ICSI tech report TR-00-03, March 2000.)195.6 411.4 Q
34.529([PFTK98] P)113.1 428 R(adh)-.165 E(ye, J. and)-.055 E(Firoiu, V)
5.5 E 2.75(.a)-1.419 G(nd T)-2.75 E -.275(ow)-.88 G(sle).275 E 1.43
-.715(y, D)-.165 H 2.75(.a).715 G(nd K)-2.75 E(urose, J., "Modeling)
-.165 E(TCP Throughput: A Simple Model and its Empirical V)195.6 441 Q
(alidation",)-1.221 E(Proc A)195.6 454 Q(CM SIGCOMM 1998.)-.44 E 29.634
([RFC2119] S.)113.1 470.6 R(Bradner)2.75 E 2.75(,K)-.44 G .33 -.165
(ey W)-3.025 H(ords F)-.715 E(or Use in RFCs to Indicate Requirement)
-.165 E(Le)195.6 483.6 Q -.165(ve)-.275 G(ls, RFC 2119.).165 E 29.634
([RFC2140] J.)113.1 500.2 R -.88(To)2.75 G
(uch, "TCP Control Block Interdependence", RFC 2140, April).88 E(1997.)
195.6 513.2 Q 29.634([RFC2581] Allman,)113.1 529.8 R(M., P)2.75 E
(axson, V)-.165 E(., and W)-1.419 E 2.75(.S)-1.012 G(te)-2.75 E -.165
(ve)-.275 G(ns, "TCP Congestion Control",).165 E(RFC 2581, April 1999.)
195.6 542.8 Q 29.634([RFC2861] M.)113.1 559.4 R(Handle)2.75 E 1.43 -.715
(y, J)-.165 H 2.75(.P).715 G(adh)-2.915 E(ye, and S. Flo)-.055 E
(yd, TCP Congestion W)-.11 E(indo)-.44 E(w)-.275 E -1.221(Va)195.6 572.4
S(lidation, RFC2861, June 2000.)1.221 E 29.634([RFC2988] V)113.1 589 R
2.75(.P)-1.419 G(axson and M. Allman, "Computing TCP')-2.915 E 2.75(sR)
-.605 G(etransmission)-2.75 E -.385(Ti)195.6 602 S(mer", RFC 2988, No)
.385 E -.165(ve)-.165 G(mber 2000.).165 E 29.634([RFC3168] K.)113.1
618.6 R(Ramakrishnan and S. Flo)2.75 E
(yd, "The Addition of Explicit Congestion)-.11 E
(Noti\214cation \(ECN\) to IP", RFC 3168, September 2001.)195.6 631.6 Q
29.634([RFC3390] Allman,)113.1 648.2 R(M., Flo)2.75 E(yd, S., and C. P)
-.11 E(artridge, "Increasing TCP')-.165 E 2.75(sI)-.605 G(nitial)-2.75 E
-.44(Wi)195.6 661.2 S(ndo).44 E(w", RFC 3390, October 2002.)-.275 E
15.587([RFC3448Err] RFC)113.1 677.8 R(3448 Errata, URL `)2.75 E
(`http://www)-.814 E(.icir)-.715 E(.or)-.605 E(g/tfrc/rfc3448.errata')
-.198 E('.)-.814 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E
(ye/W)-.055 E 247.343(idmer [P)-.44 F(age 37])-.165 E 0 Cg EP
%%Page: 38 38
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E 29.634([RFC3540] W)113.1 85 R
(etherall, D., Ely)-.88 E 2.75(,D)-.715 G(., and Spring, N., "Rob)-2.75
E(ust ECN Signaling with)-.22 E
(Nonces", RFC 3540, Experimental, June 2003)195.6 98 Q 29.634
([RFC4340] K)113.1 114.6 R(ohler)-.385 E 2.75(,E)-.44 G(., Handle)-2.75
E 1.43 -.715(y, M)-.165 H(., and S. Flo).715 E(yd, "Datagram Congestion)
-.11 E(Control Protocol \(DCCP\)", RFC 4340, March 2006.)195.6 127.6 Q
29.634([RFC4342] Flo)113.1 144.2 R(yd, S., K)-.11 E(ohler)-.385 E 2.75
(,E)-.44 G(., and J. P)-2.75 E(adh)-.165 E(ye, "Pro\214le for Datagram)
-.055 E
(Congestion Control Protocol \(DCCP\) Congestion Control ID 3: TCP-)
195.6 157.2 Q(Friendly Rate Control \(TFRC\)", RFC 4342, March 2006.)
195.6 170.2 Q 29.634([RFC4828] Flo)113.1 186.8 R(yd, S., and E. K)-.11 E
(ohler)-.385 E 2.75(,T)-.44 G(CP Friendly Rate Control \(TFRC\): the)
-2.75 E(Small-P)195.6 199.8 Q(ack)-.165 E(et \(SP\) V)-.11 E
(ariant, RFC 4828, Experimental, April 2007.)-1.221 E 51.04([W00] W)
113.1 216.4 R(idmer)-.44 E 2.75(,J)-.44 G
(., "Equation-Based Congestion Control", Diploma Thesis,)-2.75 E(Uni)
195.6 229.4 Q -.165(ve)-.275 G(rsity of Mannheim, February 2000.).165 E
(URL)5.5 E("http://www)195.6 242.4 Q(.icir)-.715 E(.or)-.605 E
(g/tfrc/".)-.198 E/F1 14/Times-Bold@0 SF -.7(Au)72 281.4 S(thors' Addr)
.7 E(esses)-.252 E F0(Mark Handle)136.8 300.4 Q -.715(y,)-.165 G
(Department of Computer Science)136.8 313.4 Q(Uni)136.8 326.4 Q -.165
(ve)-.275 G(rsity Colle).165 E(ge London)-.165 E(Go)136.8 339.4 Q
(wer Street)-.275 E(London WC1E 6BT)136.8 352.4 Q(UK)136.8 365.4 Q
(EMail: M.Handle)136.8 378.4 Q(y@cs.ucl.ac.uk)-.165 E(Sally Flo)136.8
403.4 Q(yd)-.11 E(ICSI)136.8 416.4 Q(1947 Center St, Suite 600)136.8
429.4 Q(Berk)136.8 442.4 Q(ele)-.11 E 1.43 -.715(y, C)-.165 H 2.75(A9)
.715 G(4708)-2.75 E<8d6f>136.8 455.4 Q(yd@icir)-.11 E(.or)-.605 E(g)
-.198 E(Jitendra P)136.8 480.4 Q(adh)-.165 E(ye)-.055 E
(Microsoft Research)136.8 493.4 Q(padh)136.8 506.4 Q(ye@microsoft.com)
-.055 E .918<4aa8>136.8 544.4 S(or)-5.499 E 2.75(gW)-.198 G(idmer)-3.19
E(DoCoMo Euro-Labs)136.8 557.4 Q(Landsber)136.8 570.4 Q(ger Strasse 312)
-.198 E(80687 Munich)136.8 583.4 Q(German)136.8 596.4 Q(y)-.165 E
(widmer@acm.or)136.8 609.4 Q(g)-.198 E F1(Full Copyright Statement)72
654.4 Q F0(Cop)100.8 671 Q(yright \(C\) The IETF T)-.11 E
(rust \(2007\).)-.385 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)
-.165 E(ye/W)-.055 E 247.343(idmer [P)-.44 F(age 38])-.165 E 0 Cg EP
%%Page: 39 39
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2008)2.75 E(July 2007)123.726 E(This document is subject to th\
e rights, licenses and restrictions contained in BCP 78, and)100.8 85 Q
-.165(ex)100.8 98 S
(cept as set forth therein, the authors retain all their rights.).165 E
(This document and the information contained herein are pro)100.8 114.6
Q(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)100.8 127.6 Q(UT)
-.11 E(OR, THE ORGANIZA)-.198 E(TION HE/SHE REPRESENTS OR IS)-1.221 E
(SPONSORED BY \(IF ANY\), THE INTERNET SOCIETY)100.8 140.6 Q 2.75(,T)
-1.419 G(HE IETF TR)-2.75 E(UST AND THE)-.44 E(INTERNET ENGINEERING T)
100.8 153.6 Q(ASK FORCE DISCLAIM ALL W)-1.023 E(ARRANTIES,)-1.32 E
(EXPRESS OR IMPLIED, INCLUDING B)100.8 166.6 Q(UT NO)-.11 E 2.75(TL)-.44
G(IMITED T)-2.75 E 2.75(OA)-.198 G(NY W)-2.75 E(ARRANTY)-1.32 E(THA)
100.8 179.6 Q 2.75(TT)-1.221 G(HE USE OF THE INFORMA)-2.75 E
(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G(NFRINGE ANY)-2.75 E
(RIGHTS OR ANY IMPLIED W)100.8 192.6 Q(ARRANTIES OF MERCHANT)-1.32 E
(ABILITY OR FITNESS)-1.023 E(FOR A P)100.8 205.6 Q(AR)-1.012 E
(TICULAR PURPOSE.)-.66 E/F1 14/Times-Bold@0 SF(Intellectual Pr)72 231.6
Q(operty)-.252 E F0(The IETF tak)100.8 248.2 Q(es no position re)-.11 E
-.055(ga)-.165 G(rding the v).055 E(alidity or scope of an)-.275 E 2.75
(yI)-.165 G(ntellectual Property)-2.75 E(Rights or other rights that mi\
ght be claimed to pertain to the implementation or use of the)100.8
261.2 Q(technology described in this document or the e)100.8 274.2 Q
(xtent to which an)-.165 E 2.75(yl)-.165 G(icense under such rights)
-2.75 E(might or might not be a)100.8 287.2 Q -.275(va)-.22 G
(ilable; nor does it represent that it has made an).275 E 2.75(yi)-.165
G(ndependent)-2.75 E(ef)100.8 300.2 Q(fort to identify an)-.275 E 2.75
(ys)-.165 G(uch rights.)-2.75 E
(Information on the procedures with respect to rights in)5.5 E
(RFC documents can be found in BCP 78 and BCP 79.)100.8 313.2 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)100.8
329.8 Q 2.75(ya)-.165 G(ssurances of licenses to)-2.75 E(be made a)100.8
342.8 Q -.275(va)-.22 G(ilable, or the result of an attempt made to obt\
ain a general license or).275 E(permission for the use of such propriet\
ary rights by implementers or users of this)100.8 355.8 Q(speci\214cati\
on can be obtained from the IETF on-line IPR repository at)100.8 368.8 Q
(http://www)100.8 381.8 Q(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E
(The IETF in)100.8 398.4 Q(vites an)-.44 E 2.75(yi)-.165 G
(nterested party to bring to its attention an)-2.75 E 2.75(yc)-.165 G
(op)-2.75 E(yrights, patents or)-.11 E
(patent applications, or other proprietary rights that may co)100.8
411.4 Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)100.8 424.4 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
100.8 437.4 Q(g.)-.198 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)
-.165 E(ye/W)-.055 E 247.343(idmer [P)-.44 F(age 39])-.165 E 0 Cg EP
%%Trailer
end
%%EOF