One document matched: draft-ietf-dccp-rfc3448bis-01.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.18.1
%%CreationDate: Sun Mar  4 22:15:24 2007
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.18 1
%%Pages: 30
%%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
(M. Handley)186 E 192(INTERNET-DRAFT University)72 98 R(College London)6
E(Intended status: Proposed Standard)72 111 Q(S. Floyd)180 E
(Expires: September 2007)72 124 Q(ICIR)270 E(J. Padhye)450 137 Q
(Microsoft)450 150 Q(J. Widmer)450 163 Q(University of Mannheim)372 176
Q 6(4M)432 189 S(arch 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-01.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 September 2007.)-.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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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(7)A(2. Con)100.8 122 Q -.15(ve)-.4 G
(ntions).15 E F0 -2.25(................................................\
..................................................................... .\
...)4.94 F F2(8)A(3. Protocol Mechanism)100.8 134 Q F0 -2.25(..........\
.......................................................................\
........................ ....)4.9 F F2(8)A(3.1. TCP Throughput Equation)
115.8 146 Q F0 -2.25(..................................................\
....................................... ....)2.78 F F2(8)A(3.2. P)115.8
158 Q(ack)-.15 E(et Contents)-.1 E F0 -2.25(...........................\
.......................................................................\
...... ....)3.73 F F2(9)A(3.2.1. Data P)140.8 170 Q(ack)-.15 E(ets)-.1 E
F0 -1.75(..............................................................\
................................... ..)3.27 F F2(10)A(3.2.2. Feedback P)
140.8 182 Q(ack)-.15 E(ets)-.1 E F0 -1.75(.............................\
............................................................. ..)3.08 F
F2(10)A(4. Data Sender Protocol)100.8 194 Q F0 -1.75(..................\
.......................................................................\
............... ..)5.16 F F2(10)A(4.1. Measuring the Se)115.8 206 Q
(gment Size)-.15 E F0 -1.75(...........................................\
........................................... ..)3.14 F F2(11)A
(4.2. Sender Initialization)115.8 218 Q F0 -1.75(......................\
.......................................................................\
..... ..)3.32 F F2(11)A(4.3. Sender beha)115.8 230 Q
(vior when a feedback pack)-.2 E(et is recei)-.1 E -.15(ve)-.25 G(d).15
E F0 -1.75(.................................................... ..)2.81
F F2(12)A(4.4. Expiration of nofeedback timer)115.8 242 Q F0 -1.75(....\
.......................................................................\
....... ..)3.45 F F2(13)A(4.5. Sending a pack)115.8 254 Q
(et after an idle or data-limited period)-.1 E F0 -1.75
(.................................................... ..)3.01 F F2(14)A
(4.6. Pre)115.8 266 Q -.15(ve)-.25 G(nting Oscillations).15 E F0 -1.75(\
.......................................................................\
...................... ..)5.24 F F2(14)A(4.7. Scheduling of P)115.8 278
Q(ack)-.15 E(et T)-.1 E(ransmissions)-.35 E F0 -1.75(..................\
......................................................... ..)3.28 F F2
(15)A(5. Calculation of the Loss Ev)100.8 290 Q(ent Rate \(p\))-.15 E F0
-1.75(.................................................................\
............... ..)2.98 F F2(15)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(15)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(16)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(17)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(17)A(5.5. History Discounting)115.8 350 Q F0
-1.75(.................................................................\
................................. ..)3.31 F F2(18)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(20)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 pack)-.2 E(et is recei)-.1 E
-.15(ve)-.25 G(d).15 E F0 -1.75
(........................................................ ..)3.87 F F2
(20)A(6.2. Expiration of feedback timer)115.8 386 Q F0 -1.75(..........\
.......................................................................\
.... ..)5.2 F F2(21)A(6.3. Recei)115.8 398 Q -.15(ve)-.25 G 2.5(ri).15 G
(nitialization)-2.5 E F0 -1.75(........................................\
....................................................... ..)4.75 F F2(21)
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
(22)A(7. Sender)100.8 422 Q(-based V)-.2 E(ariants)-1.11 E F0 -1.75(...\
.......................................................................\
............................. ..)4.51 F F2(23)A
(8. Implementation Issues)100.8 434 Q F0 -1.75(........................\
.......................................................................\
........ ..)3.18 F F2(23)A(9. Changes from RFC 3448)100.8 446 Q F0 -1.75
(......................................................................\
............................. ..)4.17 F F2(24)A
(10. Security Considerations)100.8 458 Q F0 -1.75(.....................\
.......................................................................\
....... ..)3.62 F F2(25)A(11. IAN)100.8 470 Q 2.5(AC)-.35 G
(onsiderations)-2.5 E F0 -1.75(........................................\
.............................................................. ..)4.06 F
F2(25)A(12. Ackno)100.8 482 Q(wledgments)-.25 E F0 -1.75(..............\
.......................................................................\
...................... ..)2.71 F F2(25)A(13. T)100.8 494 Q(erminology)
-.7 E F0 -1.75(........................................................\
........................................................... ..)3.93 F F2
(26)A(14. Normati)100.8 506 Q .3 -.15(ve R)-.25 H(eferences).15 E F0
-1.75(.................................................................\
.................................... ..)4.66 F F2(28)A
(15. Informational References)100.8 518 Q F0 -1.75(....................\
.......................................................................\
...... ..)3.04 F F2(28)A(16. Authors' Addresses)100.8 530 Q F0 -1.75(..\
.......................................................................\
................................ ..)3.8 F F2(29)A(Full Cop)100.8 542 Q
(yright Statement)-.1 E F0 -1.75(......................................\
................................................................ ..)4.62
F F2(29)A(Intellectual Property)100.8 554 Q F0 -1.75(..................\
.......................................................................\
.................... ..)5.02 F F2(30)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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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-00.txt:)103.55 111 Q 2.75(*W)
103.55 137 S(hen initializing the loss history after the \214rst)-2.75 E
(data pack)109.05 150 Q(et sent is lost or ECN-mark)-.11 E
(ed, TFRC uses)-.11 E 2.75(am)109.05 163 S(inimum recei)-2.75 E .33
-.165(ve r)-.275 H(ate of 0.5 pack).165 E(ets per second.)-.11 E 2.75
(*F)103.55 189 S(or initializing the estimated pack)-2.915 E
(et drop rate)-.11 E(for the \214rst loss interv)109.05 202 Q
(al when coming out of slo)-.275 E(w-start,)-.275 E
(it is ok to use the maximum recei)109.05 215 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)
109.05 228 Q .33 -.165(ve r)-.275 H(ate in the last round-trip time.)
.165 E(Feedback from Ladan Gharai.)109.05 241 Q 2.75(*G)103.55 267 S
(eneral feedback from Gorry F)-2.75 E(airhurst:)-.165 E 2.75(-A)109.05
280 S(dded a reference for TFRC-SP)-2.75 E(.)-1.221 E 2.75(-C)109.05 293
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.)114.55 306 Q 2.75(-A)109.05 319 S
(dded a de\214nition of terms.)-2.75 E 2.75(-A)109.05 332 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)114.55 345 Q 2.75(os)-.11 G
(econds, instead of three seconds for the)-2.75 E(recommended initial v)
114.55 358 Q(alue for TCP')-.275 E 2.75(sr)-.605 G(etransmit timer)-2.75
E(.)-.605 E 2.75(*G)103.55 384 S
(eneral feedback from Arjuna Sathiaseelan:)-2.75 E 2.75(-A)109.05 397 S
(dded more details about sending multiple feedback)-2.75 E(pack)117.3
410 Q(ets per R)-.11 E(TT)-.66 E(.)-.814 E 2.75(-A)109.05 423 S
(dded change to Section 4.3 to use the \214rst feedback)-2.75 E(pack)
117.3 436 Q(et, or the \214rst feedback pack)-.11 E(et after a)-.11 E
(nofeedback timer during slo)117.3 449 Q(w-start, *if min_rate > X*.)
-.275 E 2.75(*G)103.55 475 S(eneral feedback from Gerrit Renk)-2.75 E
(er:)-.11 E 2.75(-C)109.05 488 S(hanged "delta" to "t_delta".)-2.75 E
2.75(-C)109.05 501 S(hanged X_calc to X_Bps, clari\214ed X.)-2.75 E 2.75
(-C)109.05 514 S(lari\214ed send times in Section 4.7.)-2.75 E 2.75(-C)
109.05 527 S(hanged so that tld can be initialized to either 0 or -1.)
-2.75 E 2.75(-F)109.05 540 S(ix)-2.75 E
(ed Section 5.5 to say that the most recent lost)-.165 E(interv)114.55
553 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)114.55 566 Q(als*.)-.275 E
2.75(-C)109.05 579 S(lari\214ed introduction about \214x)-2.75 E
(ed-size and v)-.165 E(ariable-size)-.275 E(pack)114.55 592 Q(ets.)-.11
E 2.75(*A)103.55 618 S(dded more about sender)-2.75 E(-based v)-.22 E
(ariants.)-.275 E(Feedback from Guillaume Jourjon.)109.05 631 Q 2.75(*C)
103.55 657 S(orrected that the loss interv)-2.75 E
(al I_0 includes all transmitted)-.275 E(pack)109.05 670 Q
(ets, including lost and mark)-.11 E(ed pack)-.11 E
(ets \(as de\214ned in Section)-.11 E
(5.3 in the general de\214nition.\))109.05 683 Q(Email from Eddie K)5.5
E(ohler and)-.385 E(Gerrit Renk)109.05 696 Q(er)-.11 E(.)-.605 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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E 2.75(*O)103.55 85 S
(pen issue:)-2.75 E(Feedback from Ian about problems being limited by)
5.5 E(X_recv after a loss e)109.05 98 Q -.165(ve)-.275 G 2.75(nt. There)
.165 F(might not be an easy answer)2.75 E(.)-.605 E 2.75(*R)103.55 124 S
(elated open issue:)-2.75 E(Add F)5.5 E(aster Restart to RFC3448bis?)
-.165 E(Or not?)5.5 E(From Ian McDonald.)109.05 137 Q 2.75(*O)103.55 163
S(pen issue: Adopt something lik)-2.75 E 2.75(eD)-.11 G(CCP')-2.75 E
2.75(sR)-.605 G(ecei)-2.75 E .33 -.165(ve R)-.275 H(ate Length,).165 E
(instead of ignoring one feedback pack)109.05 176 Q 2.75(et? From)-.11 F
(Eddie K)2.75 E(ohler)-.385 E(.)-.605 E 2.75(*O)103.55 202 S
(pen issue: Add possible mechanisms for limited the maximum)-2.75 E -.22
(bu)109.05 215 S(rst size?).22 E(Using a tok)5.5 E(en b)-.11 E(uck)-.22
E(et size based on the)-.11 E(current rate?)109.05 228 Q(Or not?)5.5 E
(Email from Eddie K)5.5 E(ohler and Gerrit)-.385 E(Renk)109.05 241 Q(er)
-.11 E(.)-.605 E 2.75(*R)103.55 267 S(elated open issue: T)-2.75 E 2.75
(od)-.88 G(eal with idle periods and the lik)-2.75 E(e,)-.11 E
(in Section 4.7 say that t_i := max\(t_i, t_no)109.05 280 Q 2.75(w-R)
-.275 G(TT/2\), to)-3.41 E(limit b)109.05 293 Q(ursts to R)-.22 E
(TT/2 pack)-.66 E 2.75(ets? Has)-.11 F(an)2.75 E(yone implemented this?)
-.165 E(Email from Eddie K)109.05 306 Q(ohler and Ian McDonald.)-.385 E
2.75(*N)103.55 332 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)109.05 345 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\(4*R, 2*s/X\)?)109.05 358 Q
(Email discussing pros and cons from)5.5 E(Arjuna.)109.05 371 Q 2.75(*N)
103.55 397 S(ot addressed yet: Email thread on "RFC 3448, 4.4:)-2.75 E
(Modifying)5.5 E(X_recv if p = 0 at the time of last feedback".)109.05
410 Q 2.75(*T)103.55 436 S
(odo: Update Section 9 on "Changes from RFC 3448" with)-3.63 E
(changes since draft-\215o)109.05 449 Q(yd-rfc3448bis-00.txt.)-.11 E
(Changes from draft-\215o)103.55 475 Q(yd-rfc3448bis-00.txt:)-.11 E 2.75
(*N)103.55 501 S(ame change to draft-ietf-dccp-rfc3448bis-00.txt.)-2.75
E 2.75(*S)103.55 527 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)109.05 540 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)109.05 553
S 4.378 -.814(T. F).66 H(rom feedback from Dado Colussi.).814 E 2.75(*A)
103.55 579 S(dded the procedure for sending recei)-2.75 E -.165(ve)-.275
G(r).165 E(feedback pack)109.05 592 Q(ets when a coarse-grained)-.11 E
(timestamp is used. From RFC 4243.)109.05 605 Q(Changes from RFC 3448:)
103.55 631 Q 2.75(*I)103.55 657 S
(ncorporated changes in the RFC 3448 errata:)-2.75 E 5.5(-")109.05 683 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.")117.3 696 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 4])-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E(\("T)117.3 85 Q -.11(wo)-.88
G 2.75("c).11 G(hanged to "four", for consistenc)-2.75 E 2.75(yw)-.165 G
(ith the rest)-2.75 E(of the document.)117.3 98 Q(Reported by Joer)5.5 E
2.75(gW)-.198 G(idmer\).)-3.19 E 2.75(-")109.05 124 S
(If the nofeedback timer e)-2.75 E(xpires when the sender does not)-.165
E(yet ha)114.55 137 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)114.55 150 Q -.165(ve)-.275 G
.88 -.44(r, o).165 H 2.75(rw).44 G(hen p == 0,...")-2.75 E
(\(Added "or when p == 0,", reported by W)114.55 163 Q(im Heirman\).)
-.44 E 2.75(-I)109.05 189 S 2.75(nS)-2.75 G(ection 5.5, changed:)-2.75 E
(for \(i = 1 to n\) { DF_i = 1; })120.05 202 Q(to:)114.55 215 Q
(for \(i = 0 to n\) { DF_i = 1; })120.05 228 Q(Reported by Michele R.)
114.55 241 Q 2.75(*C)103.55 267 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.)109.05 280 Q
(This includes the follo)5.5 E(wing:)-.275 E 2.75(-I)109.05 306 S
(ncorporated Section 5.1 from [RFC4342], saying that)-2.75 E
(when reducing the sending rate after an idle period, don')114.55 319 Q
(t)-.198 E(reduce the sending rate belo)114.55 332 Q 2.75(wt)-.275 G
(he initial sending rate.)-2.75 E 2.75(-C)109.05 358 S
(hange for a datalimited sender:)-2.75 E
(When the sender has been datalimited, the sender doesn')114.55 371 Q(t)
-.198 E(let the recei)114.55 384 Q .33 -.165(ve r)-.275 H
(ate limit it to a sending rate less than).165 E(the initial rate.)
114.55 397 Q 2.75(-S)109.05 423 S(mall change to slo)-2.75 E(w-start:)
-.275 E(Changed so that for the \214rst feedback pack)114.55 436 Q
(et recei)-.11 E -.165(ve)-.275 G(d,).165 E
(or for the \214rst feedback pack)114.55 449 Q(et recei)-.11 E -.165(ve)
-.275 G 2.75(da).165 G(fter an idle)-2.75 E(period, the recei)114.55 462
Q .33 -.165(ve r)-.275 H(ate is not used to limit the).165 E
(sending rate.)114.55 475 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)114.55 488 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 2.75(*C)103.55 514 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)109.05 527 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)103.55 553 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)109.05 579 S
(or initializing the loss history after the \214rst loss e)-2.915 E
-.165(ve)-.275 G(nt,).165 E(either the recei)114.55 592 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)114.55 605 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)114.55 618 Q 2.75(wa)-.275 G 2.75(ne)-2.75 G
(stimate for s.)-2.75 E 2.75(-A)109.05 644 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)114.55 657 S 2.75(nS)-2.75 G
(ection 4.1 on "Measuring the Se)-2.75 E(gment Size".)-.165 E 2.75(-C)
109.05 683 S(hanged "pack)-2.75 E(et size" to "se)-.11 E(gment size".)
-.165 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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E(END OF NO)100.8 85 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 6])-.165 E 0 Cg EP
%%Page: 7 7
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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 will be speci\214ed in a later document [TFRC-SP].)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 7])-.165 E 0 Cg EP
%%Page: 8 8
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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 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, gi)-2.75 E(ving)-.275 E
(the acceptable transmit rate.)129.6 357.2 Q 24.95<8354>100.8 373.8 S
(he sender then adjusts its transmit rate to match the calculated rate.)
-24.95 E(The dynamics of TFRC are sensiti)100.8 390.4 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
403.4 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)100.8 416.4 Q
(ut it is important to understand ho)-.22 E 2.75(wt)-.275 G
(he interactions)-2.75 E(between mechanisms af)100.8 429.4 Q
(fect the dynamics of TFRC.)-.275 E F1(3.1.)72 455.4 Q/F3 13
/Times-Bold@0 SF(TCP Thr)5.5 E(oughput Equation)-.234 E F0(An)100.8 472
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.)
100.8 485 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')100.8 498 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.)100.8 511 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)100.8 524 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)100.8 537 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)100.8 550 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.)100.8 563
Q(The throughput equation we currently recommend for TFRC is a slightly\
 simpli\214ed v)100.8 579.6 Q(ersion)-.165 E
(of the throughput equation for Reno TCP from [PFTK98]. Ideally we')
100.8 592.6 Q 2.75(dp)-.55 G(refer a throughput)-2.75 E
(equation based on SA)100.8 605.6 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)100.8 618.6 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)
100.8 631.6 Q -.165(ve)-.275 G(ly minor).165 E(.)-.605 E
(The throughput equation is:)100.8 648.2 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 8])-.165 E 0 Cg EP
%%Page: 9 9
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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 2.75(si)129.6
166.8 S 2.75(st)-2.75 G(he se)-2.75 E(gment size in bytes.)-.165 E 2.75
(Ri)129.6 183.4 S 2.75(st)-2.75 G(he round trip time in seconds.)-2.75 E
2.75(pi)129.6 200 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)129.6 213 Q
(ets transmitted.)-.11 E(t_R)129.6 229.6 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)129.6 246.2 S 2.75(st)-2.75 G(he number of pack)-2.75 E
(ets ackno)-.11 E(wledged by a single TCP ackno)-.275 E(wledgement.)
-.275 E 1.76 -.88(We f)100.8 262.8 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)100.8 275.8 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)100.8 288.8
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)100.8 301.8 Q
-.198(TO)-.66 G([RFC2988].)100.8 314.8 Q(Man)100.8 331.4 Q 2.75(yc)-.165
G(urrent TCP connections use delayed ackno)-2.75 E
(wledgements, sending an)-.275 E(ackno)100.8 344.4 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.)100.8 357.4 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)100.8 370.4 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)100.8 383.4 Q
(wledgements, we recommend b = 1.)-.275 E(In future, dif)100.8 400 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)100.8 413 Q
(conformant TCP congestion control.)100.8 426 Q
(The throughput equation can also be e)100.8 442.6 Q(xpressed as)-.165 E
(X_Bps =)100.8 461.6 Q(X_pps * s ,)5.5 E
(with X_pps, the sending rate in pack)100.8 480.6 Q(ets per second, gi)
-.11 E -.165(ve)-.275 G 2.75(na).165 G(s)-2.75 E(1)194.3 499.6 Q
(X_pps =)100.8 512.6 Q
(--------------------------------------------------------)5.5 E
(R*sqrt\(2*b*p/3\) + \(t_R)122.8 525.6 Q -.198(TO)-.66 G
(*\(3*sqrt\(3*b*p/8\)*p*\(1+32*p^2\)\)\)).198 E(The parameters s \(se)
100.8 548.2 Q(gment size\), 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.)100.8 561.2 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)100.8 574.2 Q(5. In the rest of this document all data rate\
s are measured in bytes/second.)100.8 587.2 Q/F1 11/Times-Bold@0 SF
(3.2.)72 613.2 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
629.8 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)
100.8 642.8 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
655.8 Q(et formats, as these depend)-.11 E
(on the details of the transport protocol used.)100.8 668.8 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(3.2. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E/F1 11/Times-Bold@0 SF 2.75
(3.2.1. Data)72 85 R -.11(Pa)2.75 G(ck).11 E(ets)-.11 E F0
(Each data pack)100.8 101.6 Q
(et sent by the data sender contains the follo)-.11 E(wing information:)
-.275 E 24.95<8341>100.8 118.2 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)129.6 131.2 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)129.6 144.2 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)129.6 157.2 Q
(et history at the same time.)-.11 E 24.95<8341>100.8 173.8 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)129.6 186.8 Q
(et with sequence number i.)-.11 E
(The resolution of the timestamp should)5.5 E
(typically be measured in milliseconds.)129.6 199.8 Q
(This timestamp is used by the recei)129.6 212.8 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)129.6 225.8 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)129.6 238.8 Q
.33 -.165(ve t)-.22 H(imestamps of transmitted).165 E(data pack)129.6
251.8 Q(ets.)-.11 E 1.76 -.88(We n)129.6 264.8 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)129.6 277.8 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
290.8 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)129.6 303.8
Q -.165(ve)-.275 G .88 -.44(r, a).165 H(nd where the sender).44 E(sa)
129.6 316.8 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>
100.8 333.4 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)129.6 346.4 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)129.6 359.4 Q -.165
(ve)-.275 G(nt.).165 E
(The round-trip time estimate is also used by the recei)129.6 372.4 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)129.6 385.4 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)
129.6 398.4 Q -.165(ve)-.275 G(ry quarter of a).165 E
(round-trip time, as discussed abo)129.6 411.4 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.)129.6 424.4 Q F1 2.75(3.2.2. F)72 454
R(eedback P)-.275 E(ack)-.11 E(ets)-.11 E F0(Each feedback pack)100.8
470.6 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
487.2 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)129.6 500.2 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)129.6 513.2 Q(needed if the sender does not sa)129.6 526.2 Q
.33 -.165(ve t)-.22 H(imestamps of transmitted data pack).165 E(ets.)
-.11 E 24.95<8354>100.8 542.8 S
(he amount of time elapsed between the receipt of the last data pack)
-24.95 E(et at the)-.11 E(recei)129.6 555.8 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>100.8 572.4 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(ds)
.165 G(ince the last)-2.75 E(feedback report w)129.6 585.4 Q(as sent. W)
-.11 E 2.75(ed)-.88 G(enote this by X_recv)-2.75 E(.)-.715 E 24.95<8354>
100.8 602 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 F1(4.)72 631.6 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)100.8 648.2 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)100.8 661.2 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)100.8 674.2 Q
(not recei)100.8 687.2 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
(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(4. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E(is achie)100.8 85 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)100.8 101.6 T(pecify the sender).88 E(-side protocol in the follo)
-.22 E(wing steps:)-.275 E 24.95<834d>100.8 118.2 S
(easurement of the mean se)-24.95 E(gment size being sent.)-.165 E 24.95
<8354>100.8 134.8 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
151.4 S(he sender beha)-24.95 E(vior when the nofeedback timer e)-.22 E
(xpires.)-.165 E 24.95<834f>100.8 168 S(scillation pre)-24.95 E -.165
(ve)-.275 G(ntion \(optional\)).165 E 24.95<8353>100.8 184.6 S
(cheduling of transmission on non-realtime operating systems.)-24.95 E
/F1 11/Times-Bold@0 SF(4.1.)72 214.2 Q/F2 13/Times-Bold@0 SF
(Measuring the Segment Size)5.5 E F0(The parameter s \(se)100.8 230.8 Q
(gment size\) is normally kno)-.165 E(wn to an application.)-.275 E
(This may not be so in)5.5 E(tw)100.8 243.8 Q 2.75(oc)-.11 G(ases:)-2.75
E 24.95<8328>100.8 260.4 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)129.6 273.4 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)129.6 286.4 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)129.6 299.4 Q(gment size s.)-.165 E 24.95<8328>100.8
316 S(2\) The application needs to change the se)-24.95 E
(gment size rather than the number of)-.165 E(se)129.6 329 Q
(gments per second to perform congestion control.)-.165 E(This w)5.5 E
(ould normally be the)-.11 E(case with pack)129.6 342 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)129.6 355 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)129.6 368 S 2.75(yo)
.11 G 2.75(fm)-2.75 G(easuring parameters.)-2.75 E -.165(Fo)100.8 384.6
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)100.8 397.6 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)100.8 410.6 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)100.8 423.6 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)100.8 436.6 Q .33 -.165
(ve r)-.275 H(ate and the minimum sending rate, and in).165 E
(the setting of the nofeedback timer)100.8 449.6 Q(.)-.605 E
(The TFRC recei)100.8 466.2 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)100.8 479.2 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)100.8 492.2 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-)100.8 508.8 Q(SP)100.8 521.8 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 adjust\
ing the number of pack)100.8 534.8 Q(ets sent per)-.11 E(second.)100.8
547.8 Q F1(4.2.)72 586.8 Q F2(Sender Initialization)5.5 E F0
(The initial v)100.8 603.4 Q(alues for X \(the allo)-.275 E
(wed sending rate in bytes per second\) and tld \(the T)-.275 E(ime)
-.385 E(Last Doubled during slo)100.8 616.4 Q
(w-start\) 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
629.4 Q .33 -.165(ve a r)-.22 H(ound trip sample, the v).165 E
(alue of X is)-.275 E(set to 1 MSS/second \(for MSS the Maximum Se)100.8
642.4 Q(gment Size\), the nofeedback timer is set to)-.165 E -.165(ex)
100.8 655.4 S(pire after tw).165 E 2.75(os)-.11 G
(econds, and tld is set either to 0 or to -1.)-2.75 E(Upon recei)5.5 E
(ving a round trip time)-.275 E
(measurement \(e.g., after the \214rst feedback pack)100.8 668.4 Q
(et\), tld is set to the current time, and the)-.11 E(allo)100.8 681.4 Q
(wed transmit rate X is set to W_init/R, for W_init belo)-.275 E 2.75
(wf)-.275 G(rom [RFC3390]:)-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
(4.2. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(W_init = min\(4*MSS, max\(2*MSS, 4380\)\).)136.8 85 Q -.165(Fo)100.8
104 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(If the sender does ha)100.8 120.6 Q .33 -.165(ve a r)-.22 H(ound trip\
 sample when it is ready to \214rst send data \(e.g., from the).165 E
(SYN e)100.8 133.6 Q(xchange or from a pre)-.165 E
(vious connection [RFC2140]\), the initial transmit rate X is set)-.275
E(to W_init/R, and tld is set to the current time.)100.8 146.6 Q(Wh)
100.8 163.2 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)
100.8 176.2 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')100.8 189.2 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)100.8 202.2 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)100.8 215.2 Q
(ut also to retransit a pack)-.22 E(et that is assumed to)-.11 E(ha)
100.8 228.2 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)100.8 241.2 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)100.8 254.2 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)100.8 267.2 Q(alue for TCP')
-.275 E 2.75(sr)-.605 G(etransmit timer)-2.75 E(.)-.605 E/F1 11
/Times-Bold@0 SF(4.3.)72 306.2 Q/F2 13/Times-Bold@0 SF(Sender beha)5.5 E
(vior when a feedback pack)-.325 E(et is r)-.13 E(ecei)-.234 E -.13(ve)
-.13 G(d).13 E F0(The sender kno)100.8 322.8 Q(ws its current allo)-.275
E(wed sending rate, X, and maintains an estimate of the)-.275 E
(current round trip time, R, and an estimate of the timeout interv)100.8
335.8 Q(al, t_R)-.275 E -.198(TO)-.66 G(.).198 E(When a feedback pack)
100.8 361.8 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:)100.8 374.8 Q
16.887(1\) Calculate)100.8 395 R 2.75(an)2.75 G .55 -.275(ew r)-2.75 H
(ound trip sample.).275 E(R_sample = \(t_no)129.6 408 Q 2.75(w-t)-.275 G
(_recvdata\) - t_delay)-2.75 E(.)-.715 E 16.887(2\) Update)100.8 424.6 R
(the round trip time estimate:)2.75 E(If no feedback has been recei)
165.6 443.6 Q -.165(ve)-.275 G 2.75(db).165 G(efore)-2.75 E 2.75(R=R)
176.6 456.6 S(_sample;)-2.75 E(Else)165.6 469.6 Q 2.75(R=q)176.6 482.6 S
(*R + \(1-q\)*R_sample;)-2.75 E(TFRC is not sensiti)129.6 501.6 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)129.6 514.6 Q(ault v)-.11 E(alue of 0.9.)-.275 E 16.887
(3\) Update)100.8 531.2 R(the timeout interv)2.75 E(al:)-.275 E(t_R)
165.6 550.2 Q .396 -.198(TO = 4)-.66 H(*R.).198 E 16.887(4\) Update)
100.8 572.8 R(the sending rate as follo)2.75 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 12])-.165 E 0 Cg EP
%%Page: 13 13
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(If \(sender has been idle or data-limited)165.6 85 Q(within last tw)
182.1 98 Q 2.75(or)-.11 G(ound-trip times\))-2.75 E
(min_rate = max\(2*X_recv)176.6 111 Q 2.75(,W)-.715 G(_init/R\);)-2.75 E
(Else)165.6 124 Q(min_rate = 2*X_recv;)176.6 137 Q(If \(p > 0\))165.6
150 Q(Calculate X_Bps using the TCP throughput equation.)176.6 163 Q
2.75(X=m)176.6 176 S(ax\(min\(X_Bps, min_rate\), s/t_mbi\);)-2.75 E
(Else if \(\(min_rate < X\) and \(the \214rst feedback pack)165.6 189 Q
(et, or)-.11 E(the \214rst feedback pack)182.1 202 Q
(et after a nofeedback timer\)\))-.11 E(Do nothing;)173.85 215 Q
(Else if \(t_no)165.6 228 Q 2.75(w-t)-.275 G(ld >= R\))-2.75 E 2.75(X=m)
176.6 241 S(ax\(min\(2*X, min_rate\), s/R\);)-2.75 E(tld = t_no)176.6
254 Q(w;)-.275 E(The condition `)100.8 276.6 Q
(`if \(sender has been idle or data-limited within last tw)-.814 E 2.75
(or)-.11 G(ound-trip times\)')-2.75 E(')-.814 E(pre)100.8 289.6 Q -.165
(ve)-.275 G(nts an idle or data-limited sender from ha).165 E
(ving to reduce the sending rate to less than)-.22 E
(the initial sending rate as a result of limitations from a small recei)
100.8 302.6 Q .33 -.165(ve r)-.275 H 2.75(ate. The).165 F(condition)2.75
E -.814(``)100.8 315.6 S(if \(not the \214rst feedback pack).814 E
(et, and not the \214rst feedback pack)-.11 E(et after a nofeedback)-.11
E(timer\)')100.8 328.6 Q 2.75('p)-.814 G(re)-2.75 E -.165(ve)-.275 G(nt\
s a sender from reducing the sending rate in response to a feedback pac\
k).165 E(et)-.11 E(that reports the receipt of only a fe)100.8 341.6 Q
2.75(wp)-.275 G(ack)-2.75 E(ets after start-up or after an idle period.)
-.11 E(Note that if p == 0, then the sender is in slo)100.8 358.2 Q
(w-start phase, where it approximately doubles)-.275 E(the sending rate\
 each round-trip time until a loss occurs. The s/R term gi)100.8 371.2 Q
-.165(ve)-.275 G 2.75(sam).165 G(inimum)-2.75 E(sending rate during slo)
100.8 384.2 Q(w-start of one pack)-.275 E(et per R)-.11 E(TT)-.66 E 5.5
(.T)-.814 G(he parameter t_mbi is 64 seconds,)-5.5 E
(and represents the maximum inter)100.8 397.2 Q(-pack)-.22 E(et back)
-.11 E(of)-.11 E 2.75(fi)-.275 G(nterv)-2.75 E
(al in the persistent absence of)-.275 E 2.75(feedback. Thus,)100.8
410.2 R(when p > 0 the sender sends at least one pack)2.75 E(et e)-.11 E
-.165(ve)-.275 G(ry 64 seconds.).165 E 16.887(5\) Reset)100.8 426.8 R
(the nofeedback timer to e)2.75 E
(xpire after max\(4*R, 2*s/X\) seconds.)-.165 E/F1 11/Times-Bold@0 SF
(4.4.)72 452.8 Q/F2 13/Times-Bold@0 SF(Expiration of nofeedback timer)
5.5 E F0(If the nofeedback timer e)100.8 469.4 Q
(xpires, the sender should perform the follo)-.165 E(wing actions:)-.275
E 16.887(1\) Cut)100.8 486 R(the sending rate in half.)2.75 E
(If the sender has recei)5.5 E -.165(ve)-.275 G 2.75(df).165 G
(eedback from the recei)-2.75 E -.165(ve)-.275 G -.44(r,).165 G
(this is done by modifying the sender')129.6 499 Q 2.75(sc)-.605 G
(ached cop)-2.75 E 2.75(yo)-.11 G 2.75(fX)-2.75 G(_recv \(the recei)
-2.75 E .33 -.165(ve r)-.275 H(ate\).).165 E
(Because the sending rate is limited to at most twice X_recv)129.6 512 Q
2.75(,m)-.715 G(odifying X_recv)-2.75 E
(limits the current sending rate, b)129.6 525 Q(ut allo)-.22 E
(ws the sender to slo)-.275 E(w-start, doubling its)-.275 E
(sending rate each R)129.6 538 Q(TT)-.66 E 2.75(,i)-.814 G 2.75(ff)-2.75
G(eedback messages resume reporting no losses.)-2.75 E/F3 11/Courier@0
SF(If \(X_Bps > 2*X_recv\))158.4 557 Q
(X_recv = max\(X_recv/2, s/\(2*t_mbi\)\);)184.8 570 Q(Else)158.4 583 Q
(X_recv = X_Bps/4;)184.8 596 Q F0
(The term s/\(2*t_mbi\) limits the back)129.6 618.6 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 in the case).165 E(of persistent absence of feedback.)
129.6 631.6 Q 16.887(2\) The)100.8 651.8 R -.275(va)2.75 G
(lue of X must then be recalculated as described under point \(4\) abo)
.275 E -.165(ve)-.165 G(.).165 E(If the nofeedback timer e)129.6 677.8 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)129.6 690.8 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 step)-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
(4.4. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(\(1\) can be skipped, and the sending rate cut in half directly:)129.6
85 Q 2.75(X=m)171.1 104 S(ax\(X/2, s/t_mbi\))-2.75 E 16.887(3\) Restart)
100.8 139.6 R(the nofeedback timer to e)2.75 E
(xpire after max\(4*R, 2*s/X\) seconds.)-.165 E
(Note that when the sender stops sending, the recei)100.8 156.2 Q -.165
(ve)-.275 G 2.75(rw).165 G(ill stop sending feedback.)-2.75 E(When)5.5 E
(the sender')100.8 169.2 Q 2.75(sn)-.605 G(ofeedback timer e)-2.75 E
(xpires, the sender will decrease X_recv)-.165 E 5.5(.I)-.715 G 2.75(ft)
-5.5 G(he sender)-2.75 E(subsequently starts to send ag)100.8 182.2 Q
(ain, X_recv will limit the transmit rate, and a normal)-.055 E(slo)
100.8 195.2 Q
(wstart phase will occur until the transmit rate reaches X_Bps.)-.275 E
/F1 11/Times-Bold@0 SF(4.5.)72 221.2 Q/F2 13/Times-Bold@0 SF
(Sending a pack)5.5 E(et after an idle or data-limited period)-.13 E F0
(If the sender has been idle \(unable to send because there is little o\
r no data from the)100.8 237.8 Q(application\), the allo)100.8 250.8 Q
(wed sending rate could ha)-.275 E .33 -.165(ve b)-.22 H
(een reduced due to the nofeedback).165 E(timer)100.8 263.8 Q 2.75(,a)
-.44 G 2.75(ss)-2.75 G(peci\214ed in the section abo)-2.75 E -.165(ve)
-.165 G 5.5(.B).165 G(ecause the sender is al)-5.5 E -.11(wa)-.11 G
(ys restricted to sending).11 E(at most twice the recei)100.8 276.8 Q
.33 -.165(ve r)-.275 H(ate reported by the recei).165 E -.165(ve)-.275 G
.88 -.44(r, t).165 H(he sender will be limited to at most).44 E(doublin\
g its sending rate each round-trip time, until the sending rate reaches\
 the allo)100.8 289.8 Q(wed)-.275 E
(sending rate calculated by the throughput equation.)100.8 302.8 Q F1
(4.6.)72 328.8 Q F2(Pr)5.5 E -2.99 -.195(ev e)-.234 H
(nting Oscillations).195 E F0 1.76 -.88(To p)100.8 349 T(re).88 E -.165
(ve)-.275 G(nt oscillatory beha).165 E(vior in en)-.22 E
(vironments with a lo)-.44 E 2.75(wd)-.275 G -.165(eg)-2.75 G
(ree of statistical).165 E(multiple)100.8 362 Q
(xing it is useful to modify sender')-.165 E 2.75(st)-.605 G
(ransmit rate to pro)-2.75 E(vide congestion a)-.165 E -.22(vo)-.22 G
(idance).22 E(beha)100.8 375 Q
(vior by reducing the transmit rate as the queuing delay \(and hence R)
-.22 E(TT\) increases.)-.66 E -.88(To)5.5 G
(do this the sender maintains an estimate of the long-term R)100.8 388 Q
(TT and modi\214es its sending rate)-.66 E(depending on ho)100.8 401 Q
2.75(wt)-.275 G(he most recent sample of the R)-2.75 E(TT dif)-.66 E
(fers from this v)-.275 E 2.75(alue. The)-.275 F(long-)2.75 E
(term sample is R_sqmean, and is set as follo)100.8 414 Q(ws:)-.275 E
(If no feedback has been recei)136.8 433 Q -.165(ve)-.275 G 2.75(db).165
G(efore)-2.75 E(R_sqmean = sqrt\(R_sample\);)147.8 446 Q(Else)136.8 459
Q(R_sqmean = q2*R_sqmean + \(1-q2\)*sqrt\(R_sample\);)147.8 472 Q
(Thus R_sqmean gi)100.8 491 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)100.8 504 S 2.75(Ts).66 G
2.75(amples. The)-2.75 F
(constant q2 should be set similarly to q, and we recommend a v)2.75 E
(alue of)-.275 E(0.9 as the def)100.8 517 Q(ault.)-.11 E
(The sender obtains the base allo)100.8 533.6 Q
(wed transmit rate, X, from the throughput function.)-.275 E(It then)5.5
E(calculates a modi\214ed instantaneous transmit rate X_inst, as follo)
100.8 546.6 Q(ws:)-.275 E(X_inst = X * R_sqmean / sqrt\(R_sample\);)
136.8 565.6 Q(When sqrt\(R_sample\) is greater than R_sqmean then the q\
ueue is typically increasing and)100.8 588.2 Q
(so the transmit rate needs to be decreased for stable operation.)100.8
601.2 Q(Note: This modi\214cation is not al)100.8 617.8 Q -.11(wa)-.11 G
(ys strictly required, especially if the de).11 E(gree of statistical)
-.165 E(multiple)100.8 630.8 Q(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, w).165 H 2.75
(er).44 G(ecommend that it is done because it)-2.75 E(does mak)100.8
643.8 Q 2.75(eT)-.11 G(FRC beha)-2.75 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 2.75(fs)-2.75 G(tatistical multiple)
-2.75 E(xing.)-.165 E(If it is not done, we recommend using a v)100.8
656.8 Q(ery lo)-.165 E 2.75(wv)-.275 G
(alue of q, such that q is close to or)-3.025 E -.165(ex)100.8 669.8 S
(actly zero.).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(4.6. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E/F1 11/Times-Bold@0 SF(4.7.)
72 85 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 syste\
ms typically cannot schedule e)100.8 101.6 Q -.165(ve)-.275 G(nts).165 E
(precisely)100.8 114.6 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)100.8 127.6 T
(rage rate is maintained despite the coarse-grain or irre).22 E
(gular scheduling of the operating)-.165 E 2.75(system. Thus)100.8 140.6
R 2.75(at)2.75 G(ypical sending loop will calculate the correct inter)
-2.75 E(-pack)-.22 E(et interv)-.11 E(al, t_ipi, as)-.275 E(follo)100.8
153.6 Q(ws:)-.275 E(t_ipi = s/X_inst;)136.8 172.6 Q(Let t_no)100.8 191.6
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)100.8 204.6 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
223.6 Q -.715(w,)-.275 G(t_\(i+1\) = t_i + t_ipi.)142.3 236.6 Q
(The parameter t_delta allo)100.8 255.6 Q(ws a de)-.275 E(gree of \215e)
-.165 E(xibility in the send time of a pack)-.165 E 2.75(et. When)-.11 F
(the)2.75 E(application becomes idle, it requests re-scheduling for tim\
e t_i = t_\(i-1\) + t_ipi, for t_\(i-1\))100.8 268.6 Q
(the send time for the pre)100.8 281.6 Q(vious pack)-.275 E 2.75
(et. When)-.11 F(the application is re-scheduled, it checks the)2.75 E
(current time, t_no)100.8 294.6 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
(In some cases, when the nominal send time, t_i, of the ne)100.8 311.2 Q
(xt pack)-.165 E(et is calculated, it may)-.11 E
(already be the case that t_no)100.8 324.2 Q 2.75(w>t)-.275 G
(_i - t_delta.)-2.75 E(In such a case the pack)5.5 E(et should be sent)
-.11 E(immediately)100.8 337.2 Q 5.5(.T)-.715 G(hus if the operating sy\
stem has coarse timer granularity and the transmit)-5.5 E
(rate is high, then TFRC may send short b)100.8 350.2 Q(ursts of se)-.22
E -.165(ve)-.275 G(ral pack).165 E(ets separated by interv)-.11 E
(als of)-.275 E(the OS timer granularity)100.8 363.2 Q(.)-.715 E(If the\
 operating system has a scheduling timer granularity of t_gran seconds,\
 then t_delta)100.8 379.8 Q -.11(wo)100.8 392.8 S
(uld typically be set to:).11 E(t_delta = min\(t_ipi/2, t_gran/2\);)
136.8 411.8 Q(t_gran is 10ms on man)100.8 430.8 Q 2.75(yU)-.165 G
(nix systems.)-2.75 E(If t_gran is not kno)5.5 E(wn, a v)-.275 E
(alue of 10ms can be)-.275 E(safely assumed.)100.8 443.8 Q F1(5.)72
469.8 Q/F3 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)100.8 486.4
Q -.165(ve)-.275 G(nt rate is of primary).165 E
(importance for TFRC. Loss rate measurement is performed at the recei)
100.8 499.4 Q -.165(ve)-.275 G .88 -.44(r, b).165 H(ased on the).44 E
(detection of lost or mark)100.8 512.4 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
525.4 Q -.165(ve)-.275 G 2.75(rp).165 G(rotocol.)-2.75 E F1(5.1.)72
551.4 Q F2(Detection of Lost or Mark)5.5 E(ed P)-.13 E(ack)-.13 E(ets)
-.13 E F0(TFRC assumes that all pack)100.8 568 Q
(ets contain a sequence number that is incremented by one for)-.11 E
(each pack)100.8 581 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)100.8 594 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)100.8 607 Q(et that w)-.11
E(as)-.11 E 2.75(lost. If)100.8 620 R 2.75(at)2.75 G(ransport protocol \
has the requirement that it must retransmit with the original)-2.75 E
(sequence number)100.8 633 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)100.8 646 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)
100.8 662.6 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.)100.8 675.6 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)100.8 688.6 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(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 15])-.165 E 0 Cg EP
%%Page: 16 16
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E(pack)100.8 85 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 98 Q
(ut this is implementation-speci\214c.)-.22 E(The loss of a pack)100.8
114.6 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)100.8 127.6 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)100.8 140.6 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.)100.8 153.6 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 \(after NDUP)-2.75 E -.44(AC)-1.012 G(K).44 E
(subsequent pack)100.8 166.6 Q(ets arri)-.11 E -.165(ve)-.275 G
(d\) 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(record, and the recei)100.8 179.6 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)100.8 192.6 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)100.8 205.6
Q(et reordering, b)-.11 E(ut we do not specify such a mechanism here.)
-.22 E -.165(Fo)100.8 222.2 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)100.8 235.2 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.)72 274.2 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)100.8 290.8 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)100.8 303.8 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
316.8 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)100.8 329.8 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)100.8 342.8 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)100.8 355.8 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)100.8 368.8 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)100.8 381.8 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')100.8 394.8 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)100.8 411.4 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)100.8 424.4
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)100.8 437.4 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)100.8 450.4 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)100.8 463.4 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)129.6 480 Q(et.)-.11 E
(S_before is the sequence number of the last pack)129.6 496.6 Q
(et to arri)-.11 E .33 -.165(ve w)-.275 H(ith sequence number).165 E
(before S_loss.)129.6 509.6 Q
(S_after is the sequence number of the \214rst pack)129.6 526.2 Q
(et to arri)-.11 E .33 -.165(ve w)-.275 H(ith sequence number).165 E
(after S_loss.)129.6 539.2 Q(S_max is the lar)129.6 555.8 Q
(gest sequence number)-.198 E(.)-.605 E
(T_loss is the nominal estimated arri)129.6 572.4 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.)129.6 589 Q
(T_after is the reception time of S_after)129.6 605.6 Q(.)-.605 E(Note \
that T_before can either be before or after T_after due to reordering.)
100.8 622.2 Q -.165(Fo)100.8 638.8 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)100.8 651.8 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 + \( \(T_after - T_before\))136.8 670.8 Q 2.75(*\()
169.8 683.8 S(S_loss - S_before\)/\(S_after - S_before\) \);)-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(5.2. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(Note that if the sequence space wrapped between S_before and S_after)
100.8 85 Q 2.75(,t)-.44 G(hen the sequence)-2.75 E
(numbers must be modi\214ed to tak)100.8 98 Q 2.75(et)-.11 G
(his into account before performing this calculation.)-2.75 E(If)5.5 E
(the lar)100.8 111 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)100.8 124 Q(ould normally be)
-.11 E(suf)100.8 137 Q(\214cient.)-.275 E(If the lost pack)100.8 153.6 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
166.6 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)100.8 179.6 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
196.2 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)
100.8 209.2 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.)72 238.8 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)100.8 255.4 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
268.4 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)100.8 281.4 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)100.8 294.4 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)
100.8 307.4 Q(ut not including the \214rst pack)-.22 E
(et transmitted in loss interv)-.11 E(al B.)-.275 E F1(5.4.)72 346.4 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)100.8 363 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)100.8 376 Q
-.165(ve)-.275 G(nt interv).165 E(als in such a w)-.275 E(ay that the)
-.11 E(measured loss e)100.8 389 Q -.165(ve)-.275 G
(nt rate changes smoothly).165 E(.)-.715 E -.88(We)100.8 405.6 S
(ights w_0 to w_\(n-1\) are calculated as:).88 E(If \(i < n/2\))136.8
424.6 Q(w_i = 1;)145.05 437.6 Q(Else)136.8 450.6 Q
(w_i = 1 - \(i - \(n/2 - 1\)\)/\(n/2 + 1\);)145.05 463.6 Q
(Thus if n=8, the v)100.8 486.2 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)129.6 502.8 Q(The v)100.8 519.4
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')100.8 532.4 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 not be used for v)100.8 545.4 Q
(alues of n signi\214cantly greater than 8, for traf)-.275 E<8c63>-.275
E(that might compete in the global Internet with TCP)100.8 558.4 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)100.8 571.4 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)100.8 584.4 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)100.8 601 Q -.165(ve)-.22 G(rage loss interv)
.165 E(al we need to decide whether to include the)-.275 E(interv)100.8
614 Q(al since the most recent pack)-.275 E(et loss e)-.11 E -.165(ve)
-.275 G 2.75(nt. W).165 F 2.75(eo)-.88 G(nly do this if it is suf)-2.75
E(\214ciently lar)-.275 E(ge to)-.198 E(increase the a)100.8 627 Q -.165
(ve)-.22 G(rage loss interv).165 E(al.)-.275 E
(Let the most recent loss interv)100.8 643.6 Q
(als be I_0 to I_k, where I_0 is the interv)-.275 E
(al starting with the)-.275 E(most recent loss e)100.8 656.6 Q -.165(ve)
-.275 G(nt \(if there has been one\).).165 E(If there ha)5.5 E .33 -.165
(ve b)-.22 H(een at least n loss interv).165 E(als,)-.275 E
(then k is set to n; otherwise k is the maximum number of loss interv)
100.8 669.6 Q(als seen so f)-.275 E(ar)-.11 E 5.5(.W)-.605 G(e)-6.38 E
(calculate the a)100.8 682.6 Q -.165(ve)-.22 G(rage loss interv).165 E
(al I_mean is:)-.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 17])-.165 E 0 Cg EP
%%Page: 18 18
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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 interv)-.275 E
(al I_0 that represents the number of pack)-.275 E(ets sent since)-.11 E
(the be)100.8 593.8 Q(ginning of the last loss e)-.165 E -.165(ve)-.275
G 2.75(nt. The).165 F(computation of the a)2.75 E -.165(ve)-.22 G
(rage loss interv).165 E(al using the)-.275 E(discount f)100.8 606.8 Q(\
actors is a simple modi\214cation of the procedure in Section 5.4, as f\
ollo)-.11 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(5.5. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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)-.11 E 2.75(,D)-.44 G 2.75
(Fi)-2.75 G 2.75(su)-2.75 G(pdated on e)-2.75 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 number of pack).165 E
(ets sent since the)-.11 E(be)100.8 402 Q(ginning of the last loss e)
-.165 E -.165(ve)-.275 G 2.75(nt. If).165 F
(I_0 is greater than twice I_mean, then the ne)2.75 E 2.75(wl)-.275 G
(oss)-2.75 E(interv)100.8 415 Q(al is considerably lar)-.275 E
(ger than the old ones, and the general discount f)-.198 E(actor DF is)
-.11 E(updated to decrease the relati)100.8 428 Q .33 -.165(ve w)-.275 H
(eight on the older interv).165 E(als, as follo)-.275 E(ws:)-.275 E
(if \(I_0 > 2 * I_mean\) {)136.8 460 Q(DF = 2 * I_mean/I_0;)142.3 473 Q
(if \(DF < THRESHOLD\))142.3 486 Q(DF = THRESHOLD;)147.8 499 Q 2.75(}e)
136.8 512 S(lse)-2.75 E(DF = 1;)142.3 525 Q 2.75(An)100.8 544 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)100.8 557 Q 5.5(.W)-.715 G
2.75(er)-6.38 G(ecommend a THRESHOLD of 0.5.)-2.75 E(Note)5.5 E
(that with each ne)100.8 570 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 will)-.11 E(be updated.)100.8 583 Q
(When a ne)100.8 599.6 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)100.8 612.6 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)100.8 625.6 Q 5.5(.B)
-.715 G(ecause DF_i carries the discount f)-5.5 E(actor)-.11 E
(associated with loss interv)100.8 638.6 Q
(al I_i, the DF_i array has to be shifted as well. This is done as)-.275
E(follo)100.8 651.6 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 19])-.165 E 0 Cg EP
%%Page: 20 20
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 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 pack)-.325 E(et is r)
-.13 E(ecei)-.234 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\) Let)100.8 567.6 R(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 described in)-3.025
E(Section 5.)129.6 580.6 Q 16.887(3\) If)100.8 597.2 R 2.75(p>p)2.75 G
(_pre)-2.75 E 1.43 -.715(v, c)-.275 H(ause the feedback timer to e).715
E(xpire, and perform the actions described)-.165 E(in Section 6.2)129.6
610.2 Q(If p <= p_pre)129.6 626.8 Q 2.75(vn)-.275 G 2.75(oa)-2.75 G
(ction need be performed.)-2.75 E(Ho)129.6 643.4 Q(we)-.275 E -.165(ve)
-.275 G 2.75(ra).165 G 2.75(no)-2.75 G
(ptimization might check to see if the arri)-2.75 E -.275(va)-.275 G
2.75(lo).275 G 2.75(ft)-2.75 G(he pack)-2.75 E(et caused a hole)-.11 E
(in the pack)129.6 656.4 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)-.198 E(into one.)129.6
669.4 Q(If this is the case, the recei)5.5 E -.165(ve)-.275 G 2.75(rm)
.165 G(ight also send feedback immediately)-2.75 E 5.5(.T)-.715 G(he)
-5.5 E(ef)129.6 682.4 Q(fects of such an optimization are normally e)
-.275 E(xpected to be small.)-.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
(6.1. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E/F1 11/Times-Bold@0 SF(6.2.)
72 85 Q/F2 13/Times-Bold@0 SF(Expiration of feedback timer)5.5 E F0
(When the feedback timer at the recei)100.8 101.6 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)100.8 114.6 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.)-.11 E
(Let the maximum sequence number of a pack)100.8 131.2 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)-.275 E(of the R)100.8 144.2 Q
(TT measurement included in pack)-.66 E(et S_m be R_m.)-.11 E
(As described in Section 3.2.1,)5.5 E(R_m is the sender')100.8 157.2 Q
2.75(sc)-.605 G
(urrent estimate of the round trip time, reported in data pack)-2.75 E
2.75(ets. If)-.11 F(data pack)100.8 170.2 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 pre)-2.75 E(vious feedback w)-.275 E(as sent, the recei)-.11 E
-.165(ve)-.275 G(r).165 E(performs the follo)100.8 183.2 Q(wing steps:)
-.275 E 16.887(1\) Calculate)100.8 199.8 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 16.887(2\) Calculate)100.8 216.4 R(the measured recei)2.75 E .33
-.165(ve r)-.275 H(ate, X_recv).165 E 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)-2.75 E
(the pre)129.6 229.4 Q(vious R_m seconds.)-.275 E 16.887(3\) Prepare)
100.8 246 R(and send a feedback pack)2.75 E
(et containing the information described in Section)-.11 E(3.2.2)129.6
259 Q 16.887(4\) Restart)100.8 275.6 R(the feedback timer to e)2.75 E
(xpire after R_m seconds.)-.165 E(Note that rule 2\) abo)100.8 292.2 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)100.8 305.2 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
318.2 Q -.165(ve)-.275 G(nt rate, not from a limit imposed by the).165 E
(measured recei)100.8 331.2 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)100.8 347.8 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, no feedback pack)-.11 E(et)-.11 E
(is sent, and the feedback timer is restarted to e)100.8 360.8 Q
(xpire after R_m seconds.)-.165 E F1(6.3.)72 403.4 Q F2(Recei)5.5 E -.13
(ve)-.13 G 3.25(ri).13 G(nitialization)-3.25 E F0(The recei)100.8 420 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)100.8 433 Q(et be i.)-.11 E
(When the \214rst pack)100.8 449.6 Q(et is recei)-.11 E -.165(ve)-.275 G
(d:).165 E 24.95<8353>100.8 466.2 S(et p=0)-24.95 E 24.95<8353>100.8
482.8 S 2.75(et X_recv)-24.95 F 2.75(=0)2.75 G(.)-2.75 E 24.95<8350>
100.8 499.4 S(repare and send a feedback pack)-24.95 E(et.)-.11 E 24.95
<8353>100.8 516 S(et the feedback timer to e)-24.95 E
(xpire after R_i seconds.)-.165 E(If the \214rst data pack)100.8 532.6 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)
100.8 545.6 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.)100.8 558.6 Q
(If the sender is using a coarse-grained timestamp that increments e)
100.8 575.2 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
588.2 Q(wing procedure from RFC 4342)-.275 E
(is used to determine when to send feedback messages.)100.8 601.2 Q
24.95<8357>100.8 617.8 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)129.6 630.8 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)129.6 643.8 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)129.6 656.8 Q(as sent.)-.11 E
24.95<8349>100.8 673.4 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)129.6 686.4 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. \("Greater")-.11 F(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E
(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75(6.3. [P)2.75
F(age 21])-.165 E 0 Cg EP
%%Page: 22 22
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(and "greatest" are measured in circular windo)129.6 85 Q 2.75(wc)-.275
G(ounter space.\))-2.75 E/F1 11/Times-Bold@0 SF 2.75
(6.3.1. Initializing)72 124 R(the Loss History after the First Loss Ev)
2.75 E(ent)-.11 E F0(The number of pack)100.8 140.6 Q
(ets until the \214rst loss can not be used to compute the allo)-.11 E
(wed sending)-.275 E(rate directly)100.8 153.6 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)100.8 166.6 Q 2.75
(occurred. TFRC)100.8 179.6 R(approximates this tar)2.75 E
(get rate X_tar)-.198 E(get by the maximum X_recv so f)-.198 E(ar)-.11 E
2.75(,f)-.44 G(or)-2.75 E(X_recv the recei)100.8 192.6 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(\(F)5.5 E(or a TFRC sender that al)-.165
E -.11(wa)-.11 G(ys has).11 E(data to send, it is suf)100.8 205.6 Q
(\214cient to approximate the tar)-.275 E
(get rate by the most recent X_recv)-.198 E(.)-.715 E(Ho)100.8 218.6 Q
(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H(or a TFRC sender that\
 is sometimes data-limited or idle, it is best to use the).44 E
(maximum X_recv so f)100.8 231.6 Q(ar)-.11 E(.\))-.605 E(After the \214\
rst loss, instead of initializing the \214rst loss interv)100.8 257.6 Q
(al to the number of pack)-.275 E(ets sent)-.11 E
(until the \214rst loss, the TFRC recei)100.8 270.6 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)100.8 283.6 Q
(get, and uses this synthetic loss interv)-.198 E
(al to seed the loss history)-.275 E(mechanism.)100.8 296.6 Q
(TFRC does this by \214nding some v)100.8 322.6 Q
(alue p for which the throughput equation in Section 3.1)-.275 E(gi)
100.8 335.6 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)100.8 348.6 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)100.8 361.6 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)100.8 374.6
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.)100.8 387.6 Q
(\(The 5% tolerance is introduced simply because the)5.5 E
(throughput equation is dif)100.8 400.6 Q(\214cult to in)-.275 E -.165
(ve)-.44 G(rt, and we w).165 E(ant to reduce the costs of calculating p)
-.11 E(numerically)100.8 413.6 Q(.\))-.715 E
(Special care is needed for initializing the \214rst loss interv)100.8
439.6 Q(al when the \214rst data pack)-.275 E(et is lost)-.11 E(or mark)
100.8 452.6 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
(after the retransmit timer e)100.8 465.6 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)100.8 478.6 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)100.8 491.6 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)
100.8 504.6 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)100.8 517.6 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)100.8 530.6 Q(.)-1.221 E
(When the \214rst TFRC loss interv)100.8 556.6 Q
(al is null, meaning that the \214rst data pack)-.275 E
(et is lost or ECN-)-.11 E(mark)100.8 569.6 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)100.8 582.6 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)100.8 595.6 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
608.6 Q(get of 0.5/R pack)-.198 E(ets per second, for the round-trip ti\
me R, and uses this synthetic loss)-.11 E(interv)100.8 621.6 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)100.8 634.6 Q(alue for X_tar)
-.275 E(get when initializing the \214rst loss interv)-.198 E(al.)-.275
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 22])-.165 E 0 Cg
EP
%%Page: 23 23
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E/F1 11/Times-Bold@0 SF(7.)72
85 Q/F2 14/Times-Bold@0 SF(Sender)5.5 E(-based V)-.518 E(ariants)-1.288
E F0(In a sender)100.8 101.6 Q(-based v)-.22 E
(ariant of TFRC, the recei)-.275 E -.165(ve)-.275 G 2.75(rw).165 G
(ould use reliable deli)-2.86 E -.165(ve)-.275 G(ry to send).165 E
(information about pack)100.8 114.6 Q(et losses to the sender)-.11 E
2.75(,a)-.44 G(nd the sender w)-2.75 E(ould compute the pack)-.11 E(et)
-.11 E(loss rate and the acceptable transmit rate.)100.8 127.6 Q
(The main adv)100.8 153.6 Q(antages of a sender)-.275 E(-based v)-.22 E
(ariant of TFRC w)-.275 E(ould be that the sender w)-.11 E(ould)-.11 E
(not ha)100.8 166.6 Q .33 -.165(ve t)-.22 H 2.75(ot).165 G
(rust the recei)-2.75 E -.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).44 E
(requirement of reliable deli)100.8 179.6 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(,a)-.44 G(sender)100.8
192.6 Q(-based TFRC w)-.22 E(ould ha)-.11 E .33 -.165(ve m)-.22 H
(uch tighter constraints on the transport protocol in which).165 E
(it is embedded.)100.8 205.6 Q(In contrast, the recei)100.8 231.6 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)100.8 244.6 Q
(ets, and therefore does not require the reliable deli)-.11 E -.165(ve)
-.275 G(ry of feedback).165 E(pack)100.8 257.6 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)100.8 270.6 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)100.8 296.6 Q
(-based)-.22 E -.275(va)100.8 309.6 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)100.8 322.6 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)100.8 335.6 Q(ector option, allo)-1.221 E
(wing the sender to determine e)-.275 E(xactly which)-.165 E(pack)100.8
348.6 Q(ets were dropped or mark)-.11 E
(ed, and to check the information reported in the Loss)-.11 E(Interv)
100.8 361.6 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)100.8 374.6 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(dad).165 G(ata pack)-2.75 E 2.75(et. The)-.11 F
(Ack V)2.75 E(ector)-1.221 E(option allo)100.8 387.6 Q
(ws the sender to determine for itself which data pack)-.275 E
(ets were lost or ECN-)-.11 E(mark)100.8 400.6 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)100.8 413.6 Q
(erifying information reported by the recei)-.165 E -.165(ve)-.275 G
-.605(r.).165 G F1(8.)72 452.6 Q F2(Implementation Issues)5.5 E F0(This\
 document has speci\214ed the TFRC congestion control mechanism, for us\
e by)100.8 469.2 Q(applications and transport protocols.)100.8 482.2 Q
(This section mentions brie\215y some of the fe)5.5 E(w)-.275 E
(implementation issues.)100.8 495.2 Q -.165(Fo)100.8 521.2 S 2.75(rt)
.165 G(_R)-2.75 E .396 -.198(TO = 4)-.66 H
(*R and b = 1, the throughput equation in Section 3.1 can be e).198 E
(xpressed as)-.165 E(follo)100.8 534.2 Q(ws:)-.275 E(s)169.8 553.2 Q
(X_Bps =)136.8 566.2 Q(--------)5.5 E 2.75(R*f)161.55 579.2 S(\(p\))
-2.75 E(for)100.8 598.2 Q(f\(p\) =)136.8 617.2 Q
(sqrt\(2*p/3\) + \(12*sqrt\(3*p/8\) * p * \(1+32*p^2\)\).)5.5 E 2.75(At)
100.8 636.2 S(able lookup could be used for the function f\(p\).)-2.75 E
(Man)100.8 662.2 Q 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he multiplications \(e.g., q and 1-q for the round-trip time a)-2.75 E
-.165(ve)-.22 G(rage, a f).165 E(actor of 4)-.11 E
(for the timeout interv)100.8 675.2 Q(al\) 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.)100.8 688.2 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 23])-.165 E 0 Cg EP
%%Page: 24 24
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E 1.76 -.88(We n)100.8 85 T
(ote that the optional sender mechanism for pre).88 E -.165(ve)-.275 G
(nting oscillations described in).165 E
(Section 4.6 uses a square-root computation.)100.8 98 Q -.165(Fo)100.8
124 S 2.75(rt).165 G(he calculation of the nominal arri)-2.75 E -.275
(va)-.275 G 2.75(lt).275 G(ime T_loss for a lost pack)-2.75 E
(et from Section 5.2,)-.11 E(one w)100.8 137 Q
(ay to implement this that w)-.11 E(ould a)-.11 E -.22(vo)-.22 G
(id concerns about wrapped sequence space w).22 E(ould)-.11 E
(be to use the follo)100.8 150 Q(wing:)-.275 E(T_loss = T_before +)136.8
169 Q(\(T_after - T_before\) * Dist\(S_loss,)5.5 E
(S_before\)/Dist\(S_after)136.8 182 Q 2.75(,S)-.44 G(_before\))-2.75 E
(where)100.8 201 Q
(Dist\(Seqno_A, Seqno_B\) = \(Seqno_A + 2^48 - Seqno_B\) % 2^48)136.8
220 Q(The calculation of the a)100.8 252 Q -.165(ve)-.22 G
(rage loss interv).165 E(al in Section 5.4 in)-.275 E -.22(vo)-.44 G(lv)
.22 E(es multiplications by the)-.165 E
(weights w_0 to w_\(n-1\), which for n=8 are:)100.8 265 Q
(1.0, 1.0, 1.0, 1.0, 0.8, 0.6, 0.4, 0.2.)129.6 281.6 Q -.44(Wi)100.8
298.2 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)
100.8 311.2 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.)129.6 327.8 Q(The optional \
history discounting mechanism described in Section 5.5 is used in the)
100.8 344.4 Q(calculation of the a)100.8 357.4 Q -.165(ve)-.22 G
(rage loss rate.).165 E(The history discounting mechanism is in)5.5 E
-.22(vo)-.44 G -.11(ke).22 G 2.75(do).11 G(nly)-2.75 E
(when there has been an unusually long interv)100.8 370.4 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)-.275 E(operation, the discount f)100.8 383.4
Q(actor 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.)72 422.4 Q/F2 14/Times-Bold@0 SF
(Changes fr)5.5 E(om RFC 3448)-.252 E F0
(The changes from RFC 3448 are as follo)100.8 439 Q(ws:)-.275 E 24.95
<8343>100.8 455.6 S(hanges to the initial sending rate: In RFC 3448, th\
e initial sending rate w)-24.95 E(as tw)-.11 E(o)-.11 E(pack)129.6 468.6
Q(ets per round trip time.)-.11 E
(In this document, the initial sending rate can be as high)5.5 E
(as four pack)129.6 481.6 Q(ets per round trip time, follo)-.11 E
(wing RFC 3390.)-.275 E -.165(Fo)129.6 507.6 S(llo).165 E(wing Section \
5.1 from [RFC4342], this document also speci\214es that when the)-.275 E
(sending rate is reduced after an idle period, it is not reduced belo)
129.6 520.6 Q 2.75(wt)-.275 G(he initial sending)-2.75 E 2.75(rate. In)
129.6 533.6 R(addition, when the sender has been data-limited and the s\
ender is reducing)2.75 E(the allo)129.6 546.6 Q
(wed transmit rate to twice the recei)-.275 E .33 -.165(ve r)-.275 H
(ate,, the sender doesn').165 E 2.75(tr)-.198 G(educe the)-2.75 E(allo)
129.6 559.6 Q(wed transmit rate to less than the initial sending rate.)
-.275 E 2.75(Al)129.6 585.6 S(ar)-2.75 E
(ger initial sending rate is of little use if the recei)-.198 E -.165
(ve)-.275 G 2.75(rs).165 G(ends a feedback pack)-2.75 E(et)-.11 E
(after the \214rst pack)129.6 598.6 Q(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 twice the recei)129.6 611.6 Q .33 -.165(ve r)
-.275 H 2.75(ate. In).165 F(the current document, the sender)2.75 E
(does not reduce the allo)129.6 624.6 Q
(wed sending rate to at most twice the recei)-.275 E .33 -.165(ve r)
-.275 H(ate in response).165 E(to the \214rst feedback pack)129.6 637.6
Q(et.)-.11 E 24.95<8352>100.8 654.2 S(FC 3448 had contradictory te)
-24.95 E(xt about whether the sender halv)-.165 E(ed its sending rate)
-.165 E(after *tw)129.6 667.2 Q(o* round-trip times without recei)-.11 E
(ving a feedback report, or after *four*)-.275 E(round-trip times.)129.6
680.2 Q(This document clari\214es that the sender halv)5.5 E
(es its sending rate after)-.165 E(four round-trip times without recei)
129.6 693.2 Q(ving a feedback report [RFC3448Err].)-.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 24])-.165 E 0 Cg EP
%%Page: 25 25
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E 24.95<8353>100.8 85 S
(ection 4.4 w)-24.95 E(as clari\214ed to specify that on the e)-.11 E
(xpiration of the nofeedback timer)-.165 E 2.75(,i)-.44 G(f)-2.75 E 2.75
(p=0)129.6 98 S 2.75(,s)-2.75 G
(tep \(2\) applies instead of step \(1\) [RFC3448Err].)-2.75 E 24.95
<8341>100.8 114.6 S(line in Section 5.5 w)-22.2 E(as changed from `)-.11
E(`for \(i = 1 to n\) { DF_i = 1; }')-.814 E 2.75('t)-.814 G 2.75(o`)
-2.75 G(`for \(i = 0)-3.564 E(to n\) { DF_i = 1; }')129.6 127.6 Q 2.75
('[)-.814 G(RFC3448Err].)-2.75 E 24.95<8353>100.8 144.2 S(ection 5.4 w)
-24.95 E(as modi\214ed to clarify the recei)-.11 E -.165(ve)-.275 G(r')
.165 E 2.75(sc)-.605 G(alculation of the a)-2.75 E -.165(ve)-.22 G
(rage loss).165 E(interv)129.6 157.2 Q(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 24.95<8353>100.8 173.8 S(ection 4.1 w)-24.95 E
(as modi\214ed to gi)-.11 E .33 -.165(ve a s)-.275 H
(peci\214c algorithm that could be used for).165 E(estimating the a)
129.6 186.8 Q -.165(ve)-.22 G(rage se).165 E(gment size.)-.165 E/F1 11
/Times-Bold@0 SF(10.)72 212.8 Q/F2 14/Times-Bold@0 SF
(Security Considerations)5.5 E F0
(TFRC is not a transport protocol in its o)100.8 229.4 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
242.4 Q(Therefore security)5.5 E
(primarily needs to be considered in the conte)100.8 255.4 Q
(xt of a speci\214c transport protocol and its)-.165 E
(authentication mechanisms.)100.8 268.4 Q
(Congestion control mechanisms can potentially be e)100.8 285 Q
(xploited to create denial of service.)-.165 E
(This may occur through spoofed feedback.)100.8 298 Q(Thus an)5.5 E 2.75
(yt)-.165 G(ransport protocol that uses TFRC)-2.75 E(should tak)100.8
311 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)100.8 324 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(In addition, congestio\
n control mechanisms may potentially be manipulated by a greedy)100.8
340.6 Q(recei)100.8 353.6 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
366.6 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)100.8 379.6 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
392.6 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)100.8 405.6 Q(ould depend on the tr\
ansport protocol, and in particular on whether the transport)-.11 E
(protocol is reliable or unreliable.)100.8 418.6 Q 1.76 -.88(We e)100.8
435.2 T(xpect that protocols incorporating ECN with TFRC will also w)
.715 E(ant to incorporate)-.11 E(feedback from the recei)100.8 448.2 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)100.8 461.2 Q(concealment of mark)100.8 474.2 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.)100.8 487.2
Q F1(11.)72 526.2 Q F2(IAN)5.5 E 3.5(AC)-.28 G(onsiderations)-3.5 E F0
(There are no IAN)100.8 542.8 Q 2.75(Aa)-.385 G
(ctions required for this document.)-2.75 E F1(12.)72 581.8 Q F2(Ackno)
5.5 E(wledgments)-.14 E F0 1.76 -.88(We w)100.8 598.4 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)100.8 611.4 Q(Group, the Reliable Multicast T)100.8
624.4 Q(ransport W)-.385 E(orking Group, and the End-to-End Research)
-.88 E 5.5(Group. W)100.8 637.4 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
650.4 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)100.8 663.4 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)100.8 676.4 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)
100.8 689.4 Q(xtensi)-.165 E .33 -.165(ve f)-.275 H
(eedback from using the document to produce).165 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(12. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E 2.75(aw)100.8 85 S
(orking implementation.)-2.86 E/F1 11/Times-Bold@0 SF(13.)72 111 Q/F2 14
/Times-Bold@0 SF -1.288(Te)5.5 G(rminology)1.288 E F0
(This document uses the follo)100.8 127.6 Q(wing terms:)-.275 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(13. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E(DF: discount f)100.8 85 Q
(actor for a loss interv)-.11 E(al)-.275 E
(last_counter : greatest recei)100.8 111 Q -.165(ve)-.275 G 2.75(dv).165
G(alue of the windo)-3.025 E 2.75(wc)-.275 G(ounter)-2.75 E
(min_rate : minimum transmit rate)100.8 137 Q(MSS : Maximum Se)100.8 163
Q(gment Size \(constant\))-.165 E 2.75(n:n)100.8 189 S
(umber of loss interv)-2.75 E(als)-.275 E(NDUP)100.8 215 Q -.44(AC)
-1.012 G 2.75(K:n).44 G
(umber of dupacks for inferring loss \(constant\))-2.75 E
(nofeedback timer : sender)100.8 241 Q(-side timer)-.22 E 2.75(p:m)100.8
267 S(easured Loss Ev)-2.75 E(ent Rate)-.165 E(p_pre)100.8 293 Q 2.75
(v:p)-.275 G(re)-2.75 E(vious v)-.275 E(alue of p)-.275 E 2.75(q:\214)
100.8 319 S(lter constant for R)-2.75 E(TT \(constant\))-.66 E
(q2 : \214lter constant for long-term R)100.8 345 Q(TT \(constant\))-.66
E 2.75(R:e)100.8 371 S(stimated path round-trip time)-2.75 E
(R_sample : measured path R)100.8 397 Q(TT)-.66 E
(R_sqmean : estimated long-term R)100.8 423 Q(TT)-.66 E 2.75(s:n)100.8
449 S(ominal pack)-2.75 E(et size in bytes \(constant\))-.11 E 2.75(S:s)
100.8 475 S(equence number)-2.75 E(t_delta : parameter for \215e)100.8
501 Q(xibility in send time)-.165 E
(t_gran : schedular granularity \(constant\))100.8 527 Q
(t_ipi : calculated inter)100.8 553 Q(-pack)-.22 E(et interv)-.11 E
(al for sending pack)-.275 E(ets)-.11 E(t_mbi : maximum R)100.8 579 Q
.396 -.198(TO v)-.66 H(alue of TCP \(constant\))-.077 E(tld : T)100.8
605 Q(ime Last Doubled)-.385 E(t_no)100.8 631 Q 2.75(w:c)-.275 G
(urrent time)-2.75 E(t_R)100.8 657 Q .396 -.198(TO : e)-.66 H
(stimated R).198 E .396 -.198(TO o)-.66 H 2.75(fT).198 G(CP)-2.75 E 2.75
(X:a)100.8 683 S(llo)-2.75 E(wed transmit rate)-.275 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(13. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(X_Bps : calculated sending rate in bytes per second)100.8 85 Q
(X_pps : calculated sending rate in pack)100.8 111 Q(ets per second)-.11
E(X_recv : estimated recei)100.8 137 Q .33 -.165(ve r)-.275 H
(ate at the recei).165 E -.165(ve)-.275 G(r).165 E
(X_inst : instantaneous transmit rate)100.8 163 Q
(W_init : TCP initial windo)100.8 189 Q 2.75(w\()-.275 G(constant\))
-2.75 E/F1 11/Times-Bold@0 SF(14.)72 234 Q/F2 14/Times-Bold@0 SF
(Normati)5.5 E .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252 E F1(15.)72
260 Q F2(Inf)5.5 E(ormational Refer)-.35 E(ences)-.252 E F0 40.634
([BRS99] Balakrishnan,)113.1 276.6 R
(H., Rahul, H., and Seshan, S., "An Inte)2.75 E(grated)-.165 E
(Congestion Management Architecture for Internet Hosts," Proc.)195.6
289.6 Q -.44(AC)195.6 302.6 S 2.75(MS).44 G
(IGCOMM, Cambridge, MA, September 1999.)-2.75 E 30.866([FHPW00] S.)113.1
319.2 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
332.2 Q(SIGCOMM 2000.)195.6 345.2 Q 25.982([FHPW00a] S.)113.1 361.8 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: the Extended V)195.6 374.8
Q(ersion",)-1.221 E(ICSI tech report TR-00-03, March 2000.)195.6 387.8 Q
34.529([PFTK98] P)113.1 404.4 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)195.6 417.4 Q
(alidation",)-1.221 E(Proc A)195.6 430.4 Q(CM SIGCOMM 1998.)-.44 E
29.634([RFC2119] S.)113.1 447 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)195.6 460 Q -.165(ve)-.275 G(ls, RFC 2119.).165 E 29.634
([RFC2140] J.)113.1 476.6 R -.88(To)2.75 G
(uch, "TCP Control Block Interdependence", RFC 2140, April).88 E(1997.)
195.6 489.6 Q 29.634([RFC2988] V)113.1 506.2 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)195.6 519.2 S(mer", RFC 2988, No).385 E
-.165(ve)-.165 G(mber 2000.).165 E 29.634([RFC3168] K.)113.1 535.8 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.)195.6 548.8 Q
29.634([RFC3390] Allman,)113.1 565.4 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)195.6 578.4 S(ndo).44 E(w", RFC 3390, October 2002.)-.275 E
15.587([RFC3448Err] RFC)113.1 595 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 29.634([RFC3540] W)113.1 611.6 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)195.6 624.6 Q 29.634
([RFC4340] K)113.1 641.2 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.)195.6 654.2 Q
29.634([RFC4342] Flo)113.1 670.8 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-)
195.6 683.8 Q(Friendly Rate Control \(TFRC\)", RFC 4342, March 2006.)
195.6 696.8 Q(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(15. [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 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E 29.018([TFRC-SP] Flo)113.1
85 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 98 Q(ack)
-.165 E(et \(SP\) V)-.11 E(ariant, Internet draft draft-ietf-dccp-tfrc-)
-1.221 E -.22(vo)195.6 111 S(ip-07.txt, w).22 E(ork in progress, No)-.11
E -.165(ve)-.165 G(mber 2006.).165 E(Appro)5.5 E -.165(ve)-.165 G 2.75
(df).165 G(or)-2.75 E(Experimental. .)195.6 124 Q 51.04([W00] W)113.1
140.6 R(idmer)-.44 E 2.75(,J)-.44 G
(., "Equation-Based Congestion Control", Diploma Thesis,)-2.75 E(Uni)
195.6 153.6 Q -.165(ve)-.275 G(rsity of Mannheim, February 2000.).165 E
(URL)5.5 E("http://www)195.6 166.6 Q(.icir)-.715 E(.or)-.605 E
(g/tfrc/".)-.198 E/F1 11/Times-Bold@0 SF(16.)72 205.6 Q/F2 14
/Times-Bold@0 SF -.7(Au)5.5 G(thors' Addr).7 E(esses)-.252 E F0
(Mark Handle)136.8 224.6 Q -.715(y,)-.165 G
(Department of Computer Science)136.8 237.6 Q(Uni)136.8 250.6 Q -.165
(ve)-.275 G(rsity Colle).165 E(ge London)-.165 E(Go)136.8 263.6 Q
(wer Street)-.275 E(London WC1E 6BT)136.8 276.6 Q(UK)136.8 289.6 Q
(EMail: M.Handle)136.8 302.6 Q(y@cs.ucl.ac.uk)-.165 E(Sally Flo)136.8
328.6 Q(yd)-.11 E(ICIR/ICSI)136.8 341.6 Q(1947 Center St, Suite 600)
136.8 354.6 Q(Berk)136.8 367.6 Q(ele)-.11 E 1.43 -.715(y, C)-.165 H 2.75
(A9).715 G(4708)-2.75 E<8d6f>136.8 380.6 Q(yd@icir)-.11 E(.or)-.605 E(g)
-.198 E(Jitendra P)136.8 406.6 Q(adh)-.165 E(ye)-.055 E
(Microsoft Research)136.8 419.6 Q(padh)136.8 432.6 Q(ye@microsoft.com)
-.055 E .918<4aa8>136.8 471.6 S(or)-5.499 E 2.75(gW)-.198 G(idmer)-3.19
E(Lehrstuhl Praktische Informatik IV)136.8 484.6 Q(Uni)136.8 497.6 Q
-.165(ve)-.275 G -1.832(rsit \250).165 F(at Mannheim)-4.581 E 2.75(L1)
136.8 510.6 S(5, 16 - Room 415)-2.75 E(D-68131 Mannheim)136.8 523.6 Q
(German)136.8 536.6 Q(y)-.165 E(widmer@informatik.uni-mannheim.de)136.8
549.6 Q F2(Full Copyright Statement)72 594.6 Q F0(Cop)100.8 611.2 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E(This document is\
 subject to the rights, licenses and restrictions contained in BCP 78, \
and)100.8 627.8 Q -.165(ex)100.8 640.8 S
(cept as set forth therein, the authors retain all their rights.).165 E
(This document and the information contained herein are pro)100.8 657.4
Q(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)100.8 670.4 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)100.8 683.4 Q 2.75(,T)
-1.419 G(HE IETF TR)-2.75 E(UST AND THE)-.44 E(INTERNET ENGINEERING T)
100.8 696.4 Q(ASK FORCE DISCLAIM ALL W)-1.023 E(ARRANTIES,)-1.32 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 29])-.165 E 0 Cg EP
%%Page: 30 30
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 65.168(-DRAFT Expires:)-1.012 F
(September 2007)2.75 E(March 2007)106.935 E
(EXPRESS OR IMPLIED, INCLUDING B)100.8 85 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
98 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)
100.8 111 Q(ARRANTIES OF MERCHANT)-1.32 E(ABILITY OR FITNESS)-1.023 E
(FOR A P)100.8 124 Q(AR)-1.012 E(TICULAR PURPOSE.)-.66 E/F1 14
/Times-Bold@0 SF(Intellectual Pr)72 150 Q(operty)-.252 E F0
(The IETF tak)100.8 166.6 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 might be clai\
med to pertain to the implementation or use of the)100.8 179.6 Q
(technology described in this document or the e)100.8 192.6 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)100.8 205.6 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)100.8 218.6 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.)100.8 231.6 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)100.8
248.2 Q 2.75(ya)-.165 G(ssurances of licenses to)-2.75 E(be made a)100.8
261.2 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)100.8 274.2 Q(speci\214cati\
on can be obtained from the IETF on-line IPR repository at)100.8 287.2 Q
(http://www)100.8 300.2 Q(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E
(The IETF in)100.8 316.8 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
329.8 Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)100.8 342.8 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
100.8 355.8 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 30])-.165 E 0 Cg EP
%%Trailer
end
%%EOF