One document matched: draft-ietf-dccp-rfc3448bis-00.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.18.1
%%CreationDate: Fri Oct 13 13:36:48 2006
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.18 1
%%Pages: 25
%%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 186(draft-ietf-dccp-rfc3448bis-00.ps S.)72 111 R(Floyd)6 E
(Expires: April 2007)72 124 Q(ICIR)294 E(J. Padhye)450 137 Q(Microsoft)
450 150 Q(J. Widmer)450 163 Q(University of Mannheim)372 176 Q
(13 October 2006)414 189 Q/F1 14/Times-Bold@0 SF
(TCP Friendly Rate Contr)178.086 214 Q(ol \(TFRC\):)-.252 E(Pr)222.991
227 Q(otocol Speci\214cation)-.252 E/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 April 2007.)-.165 E
F2(Abstract)72 488.6 Q F3
(This document speci\214es TCP-Friendly Rate Control \(TFRC\).)100.8
505.2 Q(TFRC is a congestion)5.5 E(control mechanism for unicast \215o)
100.8 518.2 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)100.8 531.2 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)100.8 544.2 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(,m)-1.221 G
(aking it more suitable for)-2.75 E(applications such as telephon)100.8
557.2 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.)100.8 570.2 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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 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(5)A(2. Con)100.8 122 Q -.15(ve)-.4 G
(ntions).15 E F0 -2.25(................................................\
..................................................................... .\
...)4.94 F F2(5)A(3. Protocol Mechanism)100.8 134 Q F0 -2.25(..........\
.......................................................................\
........................ ....)4.9 F F2(6)A(3.1. TCP Throughput Equation)
115.8 146 Q F0 -2.25(..................................................\
....................................... ....)2.78 F F2(6)A(3.2. P)115.8
158 Q(ack)-.15 E(et Contents)-.1 E F0 -2.25(...........................\
.......................................................................\
...... ....)3.73 F F2(7)A(3.2.1. Data P)140.8 170 Q(ack)-.15 E(ets)-.1 E
F0 -2.25(..............................................................\
................................... ....)3.27 F F2(7)A
(3.2.2. Feedback P)140.8 182 Q(ack)-.15 E(ets)-.1 E F0 -2.25(..........\
.......................................................................\
......... ....)3.08 F F2(8)A(4. Data Sender Protocol)100.8 194 Q F0
-2.25(.................................................................\
....................................... ....)5.16 F F2(8)A
(4.1. Measuring the Se)115.8 206 Q(gment Size)-.15 E F0 -2.25(.........\
.......................................................................\
...... ....)3.14 F F2(9)A(4.2. Sender Initialization)115.8 218 Q F0
-2.25(.................................................................\
................................. ....)3.32 F F2(9)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(10)A
(4.4. Expiration of nofeedback timer)115.8 242 Q F0 -1.75(.............\
..................................................................... .\
.)3.45 F F2(11)A(4.5. Sending a pack)115.8 254 Q
(et after an idle period)-.1 E F0 -1.75(...............................\
........................................... ..)4.17 F F2(11)A(4.6. Pre)
115.8 266 Q -.15(ve)-.25 G(nting Oscillations).15 E F0 -1.75(..........\
.......................................................................\
............ ..)5.24 F F2(11)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(12)A
(5. Calculation of the Loss Ev)100.8 290 Q(ent Rate \(p\))-.15 E F0
-1.75(.................................................................\
............... ..)2.98 F F2(13)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(13)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(13)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(14)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(14)A(5.5. History Discounting)115.8 350 Q F0
-1.75(.................................................................\
................................. ..)3.31 F F2(15)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(17)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
(18)A(6.2. Expiration of feedback timer)115.8 386 Q F0 -1.75(..........\
.......................................................................\
.... ..)5.2 F F2(18)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(18)
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
(19)A(7. Sender)100.8 422 Q(-based V)-.2 E(ariants)-1.11 E F0 -1.75(...\
.......................................................................\
............................. ..)4.51 F F2(19)A
(8. Implementation Issues)100.8 434 Q F0 -1.75(........................\
.......................................................................\
........ ..)3.18 F F2(20)A(9. Changes from RFC 3448)100.8 446 Q F0 -1.75
(......................................................................\
............................. ..)4.17 F F2(21)A
(10. Security Considerations)100.8 458 Q F0 -1.75(.....................\
.......................................................................\
....... ..)3.62 F F2(21)A(11. IAN)100.8 470 Q 2.5(AC)-.35 G
(onsiderations)-2.5 E F0 -1.75(........................................\
.............................................................. ..)4.06 F
F2(22)A(12. Ackno)100.8 482 Q(wledgments)-.25 E F0 -1.75(..............\
.......................................................................\
...................... ..)2.71 F F2(22)A(13. Normati)100.8 494 Q .3 -.15
(ve R)-.25 H(eferences).15 E F0 -1.75(.................................\
.................................................................... ..)
4.66 F F2(22)A(14. Informational References)100.8 506 Q F0 -1.75(......\
.......................................................................\
.................... ..)3.04 F F2(22)A(15. Authors' Addresses)100.8 518
Q F0 -1.75(............................................................\
............................................. ..)3.8 F F2(23)A(Full Cop)
100.8 530 Q(yright Statement)-.1 E F0 -1.75(...........................\
.......................................................................\
.... ..)4.62 F F2(24)A(Intellectual Property)100.8 542 Q F0 -1.75(.....\
.......................................................................\
................................. ..)5.02 F F2(24)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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 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-\215o)103.55
111 Q(yd-rfc3448bis-00.txt:)-.11 E 2.75(*N)103.55 137 S
(ame change to draft-ietf-dccp-rfc3448bis-00.txt.)-2.75 E
(Changes from RFC 3448:)103.55 163 Q 2.75(*I)103.55 189 S
(ncorporated changes in the RFC 3448 errata:)-2.75 E 5.5(-")109.05 215 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.")117.3 228 Q
(\("T)117.3 241 Q -.11(wo)-.88 G 2.75("c).11 G
(hanged to "four", for consistenc)-2.75 E 2.75(yw)-.165 G(ith the rest)
-2.75 E(of the document.)117.3 254 Q(Reported by Joer)5.5 E 2.75(gW)
-.198 G(idmer\).)-3.19 E 2.75(-")109.05 280 S(If the nofeedback timer e)
-2.75 E(xpires when the sender does not)-.165 E(yet ha)114.55 293 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)114.55 306 Q -.165(ve)-.275 G .88 -.44(r, o)
.165 H 2.75(rw).44 G(hen p == 0,...")-2.75 E
(\(Added "or when p == 0,", reported by W)114.55 319 Q(im Heirman\).)
-.44 E 2.75(-I)109.05 345 S 2.75(nS)-2.75 G(ection 5.5, changed:)-2.75 E
(for \(i = 1 to n\) { DF_i = 1; })120.05 358 Q(to:)114.55 371 Q
(for \(i = 0 to n\) { DF_i = 1; })120.05 384 Q(Reported by Michele R.)
114.55 397 Q 2.75(*C)103.55 423 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.)109.05 436 Q
(This includes the follo)5.5 E(wing:)-.275 E 2.75(-I)109.05 462 S
(ncorporated Section 5.1 from [RFC4342], saying that)-2.75 E
(when reducing the sending rate after an idle period, don')114.55 475 Q
(t)-.198 E(reduce the sending rate belo)114.55 488 Q 2.75(wt)-.275 G
(he initial sending rate.)-2.75 E 2.75(-C)109.05 514 S
(hange for a datalimited sender:)-2.75 E
(When the sender has been datalimited, the sender doesn')114.55 527 Q(t)
-.198 E(let the recei)114.55 540 Q .33 -.165(ve r)-.275 H
(ate limit it to a sending rate less than).165 E(the initial rate.)
114.55 553 Q 2.75(-S)109.05 579 S(mall change to slo)-2.75 E(w-start:)
-.275 E(Changed so that for the \214rst feedback pack)114.55 592 Q
(et recei)-.11 E -.165(ve)-.275 G(d,).165 E
(or for the \214rst feedback pack)114.55 605 Q(et recei)-.11 E -.165(ve)
-.275 G 2.75(da).165 G(fter an idle)-2.75 E(period, the recei)114.55 618
Q .33 -.165(ve r)-.275 H(ate is not used to limit the).165 E
(sending rate.)114.55 631 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)114.55 644 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)103.55 670 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)109.05 683 Q -.165(ve)-.275 G
2.75(rh).165 G(as not yet seen eight loss interv)-2.75 E(als.)-.275 E
(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E
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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 2.75(*D)103.55 85 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)109.05 111 S
(or initializing the loss history after the \214rst loss e)-2.915 E
-.165(ve)-.275 G(nt,).165 E(either the recei)114.55 124 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)114.55 137 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)114.55 150 Q 2.75(wa)-.275 G 2.75(ne)-2.75 G
(stimate for s.)-2.75 E 2.75(-A)109.05 176 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)114.55 189 S 2.75(nS)-2.75 G
(ection 4.1 on "Measuring the Se)-2.75 E(gment Size".)-.165 E 2.75(-C)
109.05 215 S(hanged "pack)-2.75 E(et size" to "se)-.11 E(gment size".)
-.165 E 2.75(*S)103.55 241 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)109.05 254 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)109.05 267
S 4.378 -.814(T. F).66 H(rom feedback from Dado Colussi.).814 E 2.75(*A)
103.55 293 S(dded, the procedure for sending recei)-2.75 E -.165(ve)
-.275 G 2.75(rf).165 G(eedback pack)-2.75 E(ets when)-.11 E 2.75(ac)
109.05 306 S(oarse-grained timestamp is used.)-2.75 E(From RFC 4243.)
8.25 E 2.75(*S)103.55 332 S
(peci\214ed the procedure for initializing the loss history then)-2.75 E
(the \214rst data pack)109.05 345 Q(et sent is lost or ECN-mark)-.11 E
5.5(ed. ***)-.11 F -.198(TO)2.75 G(DO ***.).198 E(END OF NO)100.8 371 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 4])-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 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 applications that use a \214x)100.8 411.4 Q(ed se)
-.165 E(gment size, and v)-.165 E(ary their sending rate)-.275 E
(in pack)100.8 424.4 Q(ets per second in response to congestion.)-.11 E
(TFRC can also be used by applications)5.5 E(that don')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 2.75(ds).165 G -.165(eg)-2.75 G(ment size, b).165 E(ut where the se)
-.22 E(gment size v)-.165 E(aries according to the)-.275 E
(needs of the application \(e.g., video applications\).)100.8 450.4 Q
(Some applications \(e.g., some audio applications\) require a \214x)
100.8 467 Q(ed interv)-.165 E(al of time between)-.275 E(pack)100.8 480
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;)100.8 493
Q(TFRC-SP \(Small-P)100.8 506 Q(ack)-.165 E(et TFRC\) 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)100.8 519 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
532 Q(TFRC-SP will be speci\214ed in a later document.)5.5 E
(This document speci\214es TFRC as a recei)100.8 548.6 Q -.165(ve)-.275
G -.22(r-).165 G(based mechanism, with the calculation of the).22 E
(congestion control information \(i.e., the loss e)100.8 561.6 Q -.165
(ve)-.275 G(nt rate\) in the data recei).165 E -.165(ve)-.275 G 2.75(rr)
.165 G(ather in the)-2.75 E(data sender)100.8 574.6 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)100.8 587.6 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
600.6 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)100.8 613.6 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)100.8 626.6 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].)100.8 639.6
Q F1(2.)72 665.6 Q F2(Con)5.5 E -.14(ve)-.56 G(ntions).14 E F0(The k)
100.8 682.2 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)100.8 695.2 Q(T", "RECOMMENDED", "MA)-.44 E
(Y", and "OPTION)-1.155 E(AL" in this)-.385 E(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(2. [P)2.75 F(age 5])-.165 E 0 Cg EP
%%Page: 6 6
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E
(document are to be interpreted as described in RFC 2119.)100.8 85 Q/F1
11/Times-Bold@0 SF(3.)72 111 Q/F2 14/Times-Bold@0 SF(Pr)5.5 E
(otocol Mechanism)-.252 E F0 -.165(Fo)100.8 127.6 S 2.75(ri).165 G(ts c\
ongestion control mechanism, TFRC directly uses a throughput equation f\
or the)-2.75 E(allo)100.8 140.6 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
153.6 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')100.8 166.6 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)100.8 179.6 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)100.8 192.6 Q(ed pack)-.11 E
(et refers to a congestion indication from Explicit Congestion)-.11 E
(Noti\214cation \(ECN\) [RFC3168].)100.8 205.6 Q
(Generally speaking, TFRC')100.8 222.2 Q 2.75(sc)-.605 G
(ongestion control mechanism w)-2.75 E(orks as follo)-.11 E(ws:)-.275 E
24.95<8354>100.8 238.8 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
251.8 Q(.)-.605 E 24.95<8354>100.8 268.4 S(he sender also uses these fe\
edback messages to measure the round-trip time \(R)-24.95 E(TT\).)-.66 E
24.95<8354>100.8 285 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.)129.6 298 Q 24.95<8354>100.8 314.6 S
(he sender then adjusts its transmit rate to match the calculated rate.)
-24.95 E(The dynamics of TFRC are sensiti)100.8 331.2 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
344.2 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)100.8 357.2 Q
(ut it is important to understand ho)-.22 E 2.75(wt)-.275 G
(he interactions)-2.75 E(between mechanisms af)100.8 370.2 Q
(fect the dynamics of TFRC.)-.275 E F1(3.1.)72 396.2 Q/F3 13
/Times-Bold@0 SF(TCP Thr)5.5 E(oughput Equation)-.234 E F0(An)100.8
412.8 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.)
100.8 425.8 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')100.8 438.8 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.)100.8 451.8
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)100.8 464.8 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)100.8 477.8 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)100.8 490.8 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.)
100.8 503.8 Q(The throughput equation we currently recommend for TFRC i\
s a slightly simpli\214ed v)100.8 520.4 Q(ersion)-.165 E
(of the throughput equation for Reno TCP from [PFTK98]. Ideally we')
100.8 533.4 Q 2.75(dp)-.55 G(refer a throughput)-2.75 E
(equation based on SA)100.8 546.4 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)100.8 559.4 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)
100.8 572.4 Q -.165(ve)-.275 G(ly minor).165 E(.)-.605 E
(The throughput equation is:)100.8 589 Q(s)219.3 608 Q -2.75 2.75(X= -)
136.8 621 T(---------------------------------------------------------)
-2.75 E(R*sqrt\(2*b*p/3\) + \(t_R)150.55 634 Q .396 -.198(TO * \()-.66 H
(3*sqrt\(3*b*p/8\) * p * \(1+32*p^2\)\)\)).198 E(Where:)100.8 656.6 Q
2.75(Xi)129.6 673.2 S 2.75(st)-2.75 G(he transmit rate in bytes/second.)
-2.75 E(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 6])-.165 E 0 Cg
EP
%%Page: 7 7
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 2.75(si)129.6 85 S 2.75(st)
-2.75 G(he se)-2.75 E(gment size in bytes.)-.165 E 2.75(Ri)129.6 101.6 S
2.75(st)-2.75 G(he round trip time in seconds.)-2.75 E 2.75(pi)129.6
118.2 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)129.6 131.2 Q
(ets transmitted.)-.11 E(t_R)129.6 147.8 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)129.6 164.4 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)100.8 181 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)100.8 194 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)100.8 207 S
(isting TCP implementations [W00]. Another possibility w).165 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)100.8 220 Q
-.198(TO)-.66 G([RFC2988].)100.8 233 Q(Man)100.8 249.6 Q 2.75(yc)-.165 G
(urrent TCP connections use delayed ackno)-2.75 E
(wledgements, sending an)-.275 E(ackno)100.8 262.6 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.)100.8 275.6 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)100.8 288.6 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)100.8 301.6 Q
(wledgements, we recommend b = 1.)-.275 E(In future, dif)100.8 318.2 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)100.8 331.2 Q
(conformant TCP congestion control.)100.8 344.2 Q
(The throughput equation can also be e)100.8 360.8 Q(xpressed as)-.165 E
-2.75 2.75(X= X)136.8 379.8 T(_pps * s ,)-2.75 E
(with X_pps, the sending rate in pack)100.8 398.8 Q(ets per second, gi)
-.11 E -.165(ve)-.275 G 2.75(na).165 G(s)-2.75 E(1)194.3 417.8 Q
(X_pps =)100.8 430.8 Q
(--------------------------------------------------------)5.5 E
(R*sqrt\(2*b*p/3\) + \(t_R)125.55 443.8 Q -.198(TO)-.66 G
(*\(3*sqrt\(3*b*p/8\)*p*\(1+32*p^2\)\)\)).198 E(The parameters s \(se)
100.8 466.4 Q(gment size\), 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.)100.8 479.4 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)100.8 492.4 Q(5. In the rest of this document all data rate\
s are measured in bytes/second.)100.8 505.4 Q/F1 11/Times-Bold@0 SF
(3.2.)72 531.4 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)100.8 548
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)100.8 561
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)
100.8 574 Q(et formats, as these depend)-.11 E
(on the details of the transport protocol used.)100.8 587 Q F1 2.75
(3.2.1. Data)72 626 R -.11(Pa)2.75 G(ck).11 E(ets)-.11 E F0
(Each data pack)100.8 642.6 Q
(et sent by the data sender contains the follo)-.11 E(wing information:)
-.275 E 24.95<8341>100.8 659.2 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)129.6 672.2 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)129.6 685.2 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(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 186.843(idmer Section)-.44
F 2.75(3.2.1. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(pack)129.6 85 Q
(et history at the same time.)-.11 E 24.95<8341>100.8 101.6 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)129.6 114.6 Q
(et with sequence number i.)-.11 E
(The resolution of the timestamp should)5.5 E
(typically be measured in milliseconds.)129.6 127.6 Q
(This timestamp is used by the recei)129.6 140.6 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)129.6 153.6 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)129.6 166.6 Q
.33 -.165(ve t)-.22 H(imestamps of transmitted).165 E(data pack)129.6
179.6 Q(ets.)-.11 E 1.76 -.88(We n)129.6 192.6 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)129.6 205.6 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
218.6 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)129.6 231.6
Q -.165(ve)-.275 G .88 -.44(r, a).165 H(nd where the sender).44 E(sa)
129.6 244.6 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>
100.8 261.2 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)129.6 274.2 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)129.6 287.2 Q -.165
(ve)-.275 G(nt.).165 E
(The round-trip time estimate is also used by the recei)129.6 300.2 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)129.6 313.2 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)
129.6 326.2 Q -.165(ve)-.275 G(ry quarter of a).165 E
(round-trip time, as discussed abo)129.6 339.2 Q -.165(ve)-.165 G 2.75
(,t).165 G(hen the sender does not need to send its current)-2.75 E
(estimate of the round trip time.)129.6 352.2 Q/F1 11/Times-Bold@0 SF
2.75(3.2.2. F)72 394.8 R(eedback P)-.275 E(ack)-.11 E(ets)-.11 E F0
(Each feedback pack)100.8 411.4 Q(et sent by the data recei)-.11 E -.165
(ve)-.275 G 2.75(rc).165 G(ontains the follo)-2.75 E(wing information:)
-.275 E 24.95<8354>100.8 428 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)129.6 441 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(This timestamp i\
s used by the sender to estimate the round-trip time, and is only)129.6
454 Q(needed if the sender does not sa)129.6 467 Q .33 -.165(ve t)-.22 H
(imestamps of transmitted data pack).165 E(ets.)-.11 E 24.95<8354>100.8
483.6 S
(he amount of time elapsed between the receipt of the last data pack)
-24.95 E(et at the)-.11 E(recei)129.6 496.6 Q -.165(ve)-.275 G .88 -.44
(r, a).165 H(nd the generation of this feedback report. W).44 E 2.75(ed)
-.88 G(enote this by t_delay)-2.75 E(.)-.715 E 24.95<8354>100.8 513.2 S
(he rate at which the recei)-24.95 E -.165(ve)-.275 G 2.75(re).165 G
(stimates that data w)-2.75 E(as recei)-.11 E -.165(ve)-.275 G 2.75(ds)
.165 G(ince the last)-2.75 E(feedback report w)129.6 526.2 Q(as sent. W)
-.11 E 2.75(ed)-.88 G(enote this by X_recv)-2.75 E(.)-.715 E 24.95<8354>
100.8 542.8 S(he recei)-24.95 E -.165(ve)-.275 G(r').165 E 2.75(sc)-.605
G(urrent estimate of the loss e)-2.75 E -.165(ve)-.275 G(nt rate, p.)
.165 E F1(4.)72 572.4 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)100.8 589
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)100.8 602 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)100.8 615 Q
(not recei)100.8 628 Q .33 -.165(ve a f)-.275 H(eedback report for four\
 round trip times, it cuts its sending rate in half.).165 E(This)5.5 E
(is achie)100.8 641 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)100.8 657.6 T(pecify the sender).88 E(-side protocol in the follo)
-.22 E(wing steps:)-.275 E 24.95<834d>100.8 674.2 S
(easurement of the mean se)-24.95 E(gment size being sent.)-.165 E
(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(4. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 24.95<8354>100.8 85 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>100.8 101.6 S
(he sender beha)-24.95 E(vior when the nofeedback timer e)-.22 E
(xpires.)-.165 E 24.95<834f>100.8 118.2 S(scillation pre)-24.95 E -.165
(ve)-.275 G(ntion \(optional\)).165 E 24.95<8353>100.8 134.8 S
(cheduling of transmission on non-realtime operating systems.)-24.95 E
/F1 11/Times-Bold@0 SF(4.1.)72 164.4 Q/F2 13/Times-Bold@0 SF
(Measuring the Segment Size)5.5 E F0(The parameter s \(se)100.8 181 Q
(gment size\) is normally kno)-.165 E(wn to an application.)-.275 E
(This may not be so in)5.5 E(tw)100.8 194 Q 2.75(oc)-.11 G(ases:)-2.75 E
24.95<8328>100.8 210.6 S(1\) The se)-24.95 E(gment size naturally v)
-.165 E(aries depending on the data.)-.275 E(In this case, although)5.5
E(the se)129.6 223.6 Q(gment size v)-.165 E(aries, that v)-.275 E
(ariation is not coupled to the transmit rate.)-.275 E(The TFRC)5.5 E
(sender can either compute the a)129.6 236.6 Q -.165(ve)-.22 G(rage se)
.165 E(gment size or use the maximum se)-.165 E(gment)-.165 E
(size for the se)129.6 249.6 Q(gment size s.)-.165 E 24.95<8328>100.8
266.2 S(2\) The application needs to change the se)-24.95 E
(gment size rather than the number of)-.165 E(se)129.6 279.2 Q
(gments per second to perform congestion control.)-.165 E(This w)5.5 E
(ould normally be the)-.11 E(case with pack)129.6 292.2 Q
(et audio applications where a \214x)-.11 E(ed interv)-.165 E
(al of time needs to be)-.275 E(represented by each pack)129.6 305.2 Q
2.75(et. Such)-.11 F(applications need to ha)2.75 E .33 -.165(ve a c)
-.22 H(ompletely dif).165 E(ferent)-.275 E -.11(wa)129.6 318.2 S 2.75
(yo).11 G 2.75(fm)-2.75 G(easuring parameters.)-2.75 E -.165(Fo)100.8
334.8 S 2.75(rt).165 G(he \214rst class of applications where the se)
-2.75 E(gment size v)-.165 E(aries depending on the data, the)-.275 E
(sender MA)100.8 347.8 Q 2.75(Ye)-1.155 G(stimate the se)-2.75 E
(gment size s as the a)-.165 E -.165(ve)-.22 G(rage se).165 E
(gment size o)-.165 E -.165(ve)-.165 G 2.75(rt).165 G(he last four loss)
-2.75 E(interv)100.8 360.8 Q 2.75(als. The)-.275 F(sender MA)2.75 E 2.75
(Ya)-1.155 G(lso estimate the a)-2.75 E -.165(ve)-.22 G(rage se).165 E
(gment size o)-.165 E -.165(ve)-.165 G 2.75(rl).165 G(onger time)-2.75 E
(interv)100.8 373.8 Q(als, if so desired.)-.275 E
(The TFRC sender uses the se)5.5 E(gment size s in the throughput)-.165
E(equation, in the setting of the maximum recei)100.8 386.8 Q .33 -.165
(ve r)-.275 H(ate and the minimum sending rate, and in).165 E
(the setting of the nofeedback timer)100.8 399.8 Q(.)-.605 E
(The TFRC recei)100.8 416.4 Q -.165(ve)-.275 G 2.75(rm).165 G
(ay use the a)-2.75 E -.165(ve)-.22 G(rage se).165 E
(gment size s in initializing the loss history after)-.165 E
(the \214rst loss e)100.8 429.4 Q -.165(ve)-.275 G(nt, b).165 E
(ut Section 6.3.1 also gi)-.22 E -.165(ve)-.275 G 2.75(sa).165 G 2.75
(na)-2.75 G(lternate procedure that does not use the)-2.75 E -2.475 -.22
(av e)100.8 442.4 T(rage se).22 E(gment size s.)-.165 E(The second clas\
s of applications are discussed separately in a separate document on TF\
RC-)100.8 459 Q(SP)100.8 472 Q 5.5(.F)-1.221 G(or the remainder of this\
 section we assume the sender can estimate the se)-5.665 E(gment size,)
-.165 E(and that congestion control is performed by adjusting the numbe\
r of pack)100.8 485 Q(ets sent per)-.11 E(second.)100.8 498 Q F1(4.2.)72
537 Q F2(Sender Initialization)5.5 E F0(The initial v)100.8 553.6 Q
(alues for X and tld 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
566.6 Q .33 -.165(ve a r)-.22 H(ound trip sample, the v).165 E
(alue of X is)-.275 E(set to 1 pack)100.8 579.6 Q
(et/second, the nofeedback timer is set to e)-.11 E
(xpire after 2 seconds, and the tld, the)-.165 E -.385(Ti)100.8 592.6 S
(me Last Doubled during slo).385 E(w-start, is set to -1.)-.275 E
(Upon recei)5.5 E(ving a round trip time)-.275 E
(measurement \(e.g., after the \214rst feedback pack)100.8 605.6 Q
(et\), tld is set to the current time, and the)-.11 E
(transmit rate X is set to W_init/R, for W_init belo)100.8 618.6 Q 2.75
(wf)-.275 G(rom [RFC3390]:)-2.75 E
(W_init = min\(4*MSS, max\(2*MSS, 4380\)\),)136.8 637.6 Q
(for MSS the Maximum Se)100.8 656.6 Q(gment Size.)-.165 E -.165(Fo)5.5 G
2.75(rr).165 G(esponding to the initial feedback pack)-2.75 E(et, this)
-.11 E(replaces step \(4\) of Section 4.3 belo)100.8 669.6 Q -.715(w.)
-.275 G(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(4.2. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(If the sender does ha)100.8 85
Q .33 -.165(ve a r)-.22 H(ound trip sample when it is ready to \214rst \
send data \(e.g., from the).165 E(SYN e)100.8 98 Q
(xchange or from a pre)-.165 E
(vious connection [RFC2140]\), the initial transmit rate X is set)-.275
E(to W_init/R, and tld is set to the current time.)100.8 111 Q/F1 11
/Times-Bold@0 SF(4.3.)72 137 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)100.8 153.6 Q(ws its current sending r\
ate, X, and maintains an estimate of the current)-.275 E
(round trip time, R, and an estimate of the timeout interv)100.8 166.6 Q
(al, t_R)-.275 E -.198(TO)-.66 G(.).198 E(When a feedback pack)100.8
192.6 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:)100.8 205.6 Q 16.887
(1\) Calculate)100.8 225.8 R 2.75(an)2.75 G .55 -.275(ew r)-2.75 H
(ound trip sample.).275 E(R_sample = \(t_no)129.6 238.8 Q 2.75(w-t)-.275
G(_recvdata\) - t_delay)-2.75 E(.)-.715 E 16.887(2\) Update)100.8 255.4
R(the round trip time estimate:)2.75 E(If no feedback has been recei)
165.6 274.4 Q -.165(ve)-.275 G 2.75(db).165 G(efore)-2.75 E 2.75(R=R)
176.6 287.4 S(_sample;)-2.75 E(Else)165.6 300.4 Q 2.75(R=q)176.6 313.4 S
(*R + \(1-q\)*R_sample;)-2.75 E(TFRC is not sensiti)129.6 332.4 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)129.6 345.4 Q(ault v)-.11 E(alue of 0.9.)-.275 E 16.887
(3\) Update)100.8 362 R(the timeout interv)2.75 E(al:)-.275 E(t_R)165.6
381 Q .396 -.198(TO = 4)-.66 H(*R.).198 E 16.887(4\) Update)100.8 403.6
R(the sending rate as follo)2.75 E(ws:)-.275 E
(If \(sender has been idle or data-limited\))165.6 422.6 Q
(min_rate = max\(2*X_recv)176.6 435.6 Q 2.75(,W)-.715 G(_init/R\);)-2.75
E(Else)165.6 448.6 Q(min_rate = 2*X_recv;)176.6 461.6 Q(If \(p > 0\))
165.6 474.6 Q(Calculate X_calc using the TCP throughput equation.)176.6
487.6 Q 2.75(X=m)176.6 500.6 S(ax\(min\(X_calc, min_rate\), s/t_mbi\);)
-2.75 E(Else if \(not the \214rst feedback pack)165.6 513.6 Q(et, and)
-.11 E(not the \214rst feedback pack)171.1 526.6 Q
(et after a nofeedback timer\))-.11 E(If \(t_no)176.6 539.6 Q 2.75(w-t)
-.275 G(ld >= R\))-2.75 E 2.75(X=m)187.6 552.6 S
(ax\(min\(2*X, min_rate\), s/R\);)-2.75 E(tld = t_no)187.6 565.6 Q(w;)
-.275 E(The condition `)100.8 588.2 Q
(`if \(sender has been idle or data-limited\)')-.814 E 2.75('p)-.814 G
(re)-2.75 E -.165(ve)-.275 G(nts an idle or data-limited).165 E
(sender from ha)100.8 601.2 Q(ving to reduce the sending rate to less t\
han the initial sending rate as a result)-.22 E
(of limitations from a small recei)100.8 614.2 Q .33 -.165(ve r)-.275 H
2.75(ate. The).165 F(condition `)2.75 E
(`if \(not the \214rst feedback pack)-.814 E(et,)-.11 E
(and not the \214rst feedback pack)100.8 627.2 Q
(et after a nofeedback timer\)')-.11 E 2.75('p)-.814 G(re)-2.75 E -.165
(ve)-.275 G(nts a sender from).165 E
(reducing the sending rate in response to a feedback pack)100.8 640.2 Q
(et that reports the receipt of only a)-.11 E(fe)100.8 653.2 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)100.8 669.8 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)100.8 682.8 Q
-.165(ve)-.275 G 2.75(sam).165 G(inimum)-2.75 E(sending rate during slo)
100.8 695.8 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(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 10])-.165 E 0 Cg EP
%%Page: 11 11
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E
(and represents the maximum inter)100.8 85 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 98 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)100.8 114.6 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.)72 140.6 Q/F2 13/Times-Bold@0 SF(Expiration of nofeedback timer)
5.5 E F0(If the nofeedback timer e)100.8 157.2 Q
(xpires, the sender should perform the follo)-.165 E(wing actions:)-.275
E 16.887(1\) Cut)100.8 173.8 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')129.6 186.8 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)129.6 199.8
Q 2.75(,m)-.715 G(odifying X_recv)-2.75 E
(limits the current sending rate, b)129.6 212.8 Q(ut allo)-.22 E
(ws the sender to slo)-.275 E(w-start, doubling its)-.275 E
(sending rate each R)129.6 225.8 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 \(X_calc > 2*X_recv\))158.4 244.8 Q
(X_recv = max\(X_recv/2, s/\(2*t_mbi\)\);)184.8 257.8 Q(Else)158.4 270.8
Q(X_recv = X_calc/4;)184.8 283.8 Q F0
(The term s/\(2*t_mbi\) limits the back)129.6 306.4 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.)
129.6 319.4 Q 16.887(2\) The)100.8 339.6 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)129.6 365.6 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)129.6 378.6 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
(\(1\) can be skipped, and the sending rate cut in half directly:)129.6
391.6 Q 2.75(X=m)171.1 410.6 S(ax\(X/2, s/t_mbi\))-2.75 E 16.887
(3\) Restart)100.8 446.2 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)100.8 462.8 Q -.165
(ve)-.275 G 2.75(rw).165 G(ill stop sending feedback.)-2.75 E(When)5.5 E
(the sender')100.8 475.8 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)100.8 488.8 Q
(ain, X_recv will limit the transmit rate, and a normal)-.055 E(slo)
100.8 501.8 Q
(wstart phase will occur until the transmit rate reaches X_calc.)-.275 E
F1(4.5.)72 527.8 Q F2(Sending a pack)5.5 E(et after an idle period)-.13
E F0(If the sender has been idle \(unable to send because there is no d\
ata from the application\),)100.8 544.4 Q(the allo)100.8 557.4 Q
(wed sending rate could ha)-.275 E .33 -.165(ve b)-.22 H
(een reduced due to the nofeedback timer).165 E 2.75(,a)-.44 G 2.75(ss)
-2.75 G(peci\214ed)-2.75 E(in the section abo)100.8 570.4 Q -.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 at most twice the).11 E(recei)100.8 583.4 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 doubling its)
.44 E(sending rate each round-trip time, until the sending rate reaches\
 the allo)100.8 596.4 Q(wed sending rate)-.275 E
(calculated by the throughput equation.)100.8 609.4 Q F1(4.6.)72 635.4 Q
F2(Pr)5.5 E -2.99 -.195(ev e)-.234 H(nting Oscillations).195 E F0 1.76
-.88(To p)100.8 655.6 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)100.8 668.6 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)100.8 681.6 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)100.8 694.6
Q(TT and modi\214es its sending rate)-.66 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 11])-.165 E 0 Cg EP
%%Page: 12 12
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(depending on ho)100.8 85 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 98 Q(ws:)-.275 E
(If no feedback has been recei)136.8 117 Q -.165(ve)-.275 G 2.75(db).165
G(efore)-2.75 E(R_sqmean = sqrt\(R_sample\);)147.8 130 Q(Else)136.8 143
Q(R_sqmean = q2*R_sqmean + \(1-q2\)*sqrt\(R_sample\);)147.8 156 Q
(Thus R_sqmean gi)100.8 175 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)100.8 188 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)100.8 201 Q(ault.)-.11 E(The sender obt\
ains the base transmit rate, X, from the throughput function.)100.8
217.6 Q(It then)5.5 E
(calculates a modi\214ed instantaneous transmit rate X_inst, as follo)
100.8 230.6 Q(ws:)-.275 E(X_inst = X * R_sqmean / sqrt\(R_sample\);)
136.8 249.6 Q(When sqrt\(R_sample\) is greater than R_sqmean then the q\
ueue is typically increasing and)100.8 272.2 Q
(so the transmit rate needs to be decreased for stable operation.)100.8
285.2 Q(Note: This modi\214cation is not al)100.8 301.8 Q -.11(wa)-.11 G
(ys strictly required, especially if the de).11 E(gree of statistical)
-.165 E(multiple)100.8 314.8 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
327.8 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
340.8 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)100.8 353.8 S
(actly zero.).165 E/F1 11/Times-Bold@0 SF(4.7.)72 379.8 Q/F2 13
/Times-Bold@0 SF(Scheduling of P)5.5 E(ack)-.13 E(et T)-.13 E
(ransmissions)-.962 E F0(As TFRC is rate-based, and as operating system\
s typically cannot schedule e)100.8 396.4 Q -.165(ve)-.275 G(nts).165 E
(precisely)100.8 409.4 Q 2.75(,i)-.715 G 2.75(ti)-2.75 G 2.75(sn)-2.75 G
(ecessary to be opportunistic about sending data pack)-2.75 E
(ets so that the correct)-.11 E -2.475 -.22(av e)100.8 422.4 T
(rage rate is maintained despite the coarse-grain or irre).22 E
(gular scheduling of the operating)-.165 E 2.75(system. Thus)100.8 435.4
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
448.4 Q(ws:)-.275 E(t_ipi = s/X_inst;)136.8 467.4 Q(When a sender \214r\
st starts sending at time t_0, it calculates t_ipi, and calculates a no\
minal)100.8 486.4 Q(send time t_1 = t_0 + t_ipi for pack)100.8 499.4 Q
(et 1.)-.11 E(When the application becomes idle, it checks the)5.5 E
(current time, t_no)100.8 512.4 Q 1.43 -.715(w, a)-.275 H
(nd then requests re-scheduling after \(t_ipi - \(t_no).715 E 2.75(w-t)
-.275 G(_0\)\) seconds.)-2.75 E
(When the application is re-scheduled, it checks the current time, t_no)
100.8 525.4 Q 1.43 -.715(w, a)-.275 H -.055(ga).715 G(in. If \(t_no).055
E 2.75(w>)-.275 G(t_1 - delta\) then pack)100.8 538.4 Q(et 1 is sent.)
-.11 E(No)100.8 555 Q 2.75(wan)-.275 G .55 -.275(ew t)-2.75 H(_ipi may \
be calculated, and used to calculate a nominal send time t_2 for).275 E
(pack)100.8 568 Q(et 2: t2 = t_1 + t_ipi.)-.11 E
(The process then repeats, with each successi)5.5 E .33 -.165(ve p)-.275
H(ack).165 E(et')-.11 E 2.75(ss)-.605 G(end)-2.75 E
(time being calculated from the nominal send time of the pre)100.8 581 Q
(vious pack)-.275 E(et.)-.11 E
(In some cases, when the nominal send time, t_i, of the ne)100.8 597.6 Q
(xt pack)-.165 E(et is calculated, it may)-.11 E
(already be the case that t_no)100.8 610.6 Q 2.75(w>t)-.275 G
(_i - delta.)-2.75 E(In such a case the pack)5.5 E(et should be sent)
-.11 E(immediately)100.8 623.6 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)100.8 636.6 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)100.8 649.6 Q(.)-.715 E
(The parameter delta is to allo)100.8 666.2 Q 2.75(wad)-.275 G -.165(eg)
-2.75 G(ree of \215e).165 E(xibility in the send time of a pack)-.165 E
2.75(et. If)-.11 F(the)2.75 E(operating system has a scheduling timer g\
ranularity of t_gran seconds, then delta w)100.8 679.2 Q(ould)-.11 E
(typically be set to:)100.8 692.2 Q(Handle)72 733 Q(y/Flo)-.165 E(yd/P)
-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75
(4.7. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E
(delta = min\(t_ipi/2, t_gran/2\);)136.8 85 Q(t_gran is 10ms on man)
100.8 104 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
117 Q/F1 11/Times-Bold@0 SF(5.)72 143 Q/F2 14/Times-Bold@0 SF
(Calculation of the Loss Ev)5.5 E(ent Rate \(p\))-.14 E F0
(Obtaining an accurate and stable measurement of the loss e)100.8 159.6
Q -.165(ve)-.275 G(nt rate is of primary).165 E
(importance for TFRC. Loss rate measurement is performed at the recei)
100.8 172.6 Q -.165(ve)-.275 G .88 -.44(r, b).165 H(ased on the).44 E
(detection of lost or mark)100.8 185.6 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
198.6 Q -.165(ve)-.275 G 2.75(rp).165 G(rotocol.)-2.75 E F1(5.1.)72
224.6 Q/F3 13/Times-Bold@0 SF(Detection of Lost or Mark)5.5 E(ed P)-.13
E(ack)-.13 E(ets)-.13 E F0(TFRC assumes that all pack)100.8 241.2 Q
(ets contain a sequence number that is incremented by one for)-.11 E
(each pack)100.8 254.2 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)100.8 267.2 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)100.8 280.2 Q(et that w)
-.11 E(as)-.11 E 2.75(lost. If)100.8 293.2 R 2.75(at)2.75 G(ransport pr\
otocol has the requirement that it must retransmit with the original)
-2.75 E(sequence number)100.8 306.2 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)100.8 319.2 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)100.8 335.8 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.)100.8 348.8 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)100.8 361.8 Q(ets that ha)-.11 E .33 -.165
(ve a)-.22 H(rri).165 E -.165(ve)-.275 G 2.75(da).165 G
(long with the recei)-2.75 E -.165(ve)-.275 G 2.75(rt).165 G
(imestamp when each)-2.75 E(pack)100.8 374.8 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)100.8 387.8 Q
(ut this is implementation-speci\214c.)-.22 E(The loss of a pack)100.8
404.4 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)100.8 417.4 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)100.8 430.4 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.)100.8 443.4 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 \(after NDUP)-2.75 E -.44(AC)-1.012 G(K).44 E
(subsequent pack)100.8 456.4 Q(ets arri)-.11 E -.165(ve)-.275 G
(d\) 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)100.8 469.4 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)100.8 482.4 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)100.8 495.4
Q(et reordering, b)-.11 E(ut we do not specify such a mechanism here.)
-.22 E -.165(Fo)100.8 512 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)100.8 525 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(5.2.)72 564 Q F3
-.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)100.8 580.6
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 where)-.11 E(those pack)100.8 593.6 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 \(typically\) only)-2.75 E
(performs one halving of the congestion windo)100.8 606.6 Q 2.75(wd)
-.275 G(uring an)-2.75 E 2.75(ys)-.165 G(ingle R)-2.75 E(TT)-.66 E 5.5
(.T)-.814 G(hus the recei)-5.5 E -.165(ve)-.275 G(r).165 E
(needs to map the pack)100.8 619.6 Q(et loss history into a loss e)-.11
E -.165(ve)-.275 G(nt record, where a loss e).165 E -.165(ve)-.275 G
(nt is one or).165 E(more pack)100.8 632.6 Q(ets lost in an R)-.11 E(TT)
-.66 E 5.5(.T)-.814 G 2.75(op)-6.38 G(erform this mapping, the recei)
-2.75 E -.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)-.66 E
(to use, and this is supplied periodically by the sender)100.8 645.6 Q
2.75(,t)-.44 G(ypically as control information)-2.75 E(piggy-back)100.8
658.6 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 2.75(oh).165 G .55 -.275(ow t)-2.75 H
(he R).275 E(TT measurement sent)-.66 E(to the recei)100.8 671.6 Q -.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 sender')-.22 E 2.75(sc)-.605 G(alculated R)
-2.75 E(TT)-.66 E 2.75(,R)-.814 G 2.75(,\()-2.75 G(see)-2.75 E
(Section 4.3\) for this purpose.)100.8 684.6 Q(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593(idmer Section)-.44
F 2.75(5.2. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 1.76 -.88(To d)100.8 85 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 98 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)100.8 111 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)100.8 124 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)100.8 137 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)129.6 153.6 Q(et.)-.11 E
(S_before is the sequence number of the last pack)129.6 170.2 Q
(et to arri)-.11 E .33 -.165(ve w)-.275 H(ith sequence number).165 E
(before S_loss.)129.6 183.2 Q
(S_after is the sequence number of the \214rst pack)129.6 199.8 Q
(et to arri)-.11 E .33 -.165(ve w)-.275 H(ith sequence number).165 E
(after S_loss.)129.6 212.8 Q
(T_before is the reception time of S_before.)129.6 229.4 Q
(T_after is the reception time of S_after)129.6 246 Q(.)-.605 E(Note th\
at T_before can either be before or after T_after due to reordering.)
100.8 262.6 Q -.165(Fo)100.8 279.2 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)100.8 292.2 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 + \( \(T_after - T_before\))136.8 311.2 Q 2.75(*\()
169.8 324.2 S(S_loss - S_before\)/\(S_after - S_before\) \);)-2.75 E
(Note that if the sequence space wrapped between S_before and S_after)
100.8 346.8 Q 2.75(,t)-.44 G(hen the sequence)-2.75 E
(numbers must be modi\214ed to tak)100.8 359.8 Q 2.75(et)-.11 G
(his into account before performing this calculation.)-2.75 E(If)5.5 E
(the lar)100.8 372.8 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)100.8 385.8 Q(ould normally be)
-.11 E(suf)100.8 398.8 Q(\214cient.)-.275 E(If the lost pack)100.8 415.4
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
428.4 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)100.8 441.4 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
458 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)
100.8 471 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.)72 500.6 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)100.8 517.2 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
530.2 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)100.8 543.2 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 F1(5.4.)72 595.2 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)100.8 611.8 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)100.8 624.8 Q
-.165(ve)-.275 G(nt interv).165 E(als in such a w)-.275 E(ay that the)
-.11 E(measured loss e)100.8 637.8 Q -.165(ve)-.275 G
(nt rate changes smoothly).165 E(.)-.715 E -.88(We)100.8 654.4 S
(ights w_0 to w_\(n-1\) are calculated as:).88 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 14])-.165 E 0 Cg EP
%%Page: 15 15
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(If \(i < n/2\))136.8 85 Q
(w_i = 1;)145.05 98 Q(Else)136.8 111 Q
(w_i = 1 - \(i - \(n/2 - 1\)\)/\(n/2 + 1\);)145.05 124 Q
(Thus if n=8, the v)100.8 146.6 Q(alues of w_0 to w_7 are:)-.275 E
(1.0, 1.0, 1.0, 1.0, 0.8, 0.6, 0.4, 0.2)129.6 163.2 Q(The v)100.8 179.8
Q(alue n for the number of loss interv)-.275 E
(als used in calculating the loss e)-.275 E -.165(ve)-.275 G(nt rate)
.165 E(determines TFRC')100.8 192.8 Q 2.75(ss)-.605 G
(peed in responding to changes in the le)-2.75 E -.165(ve)-.275 G 2.75
(lo).165 G 2.75(fc)-2.75 G 2.75(ongestion. As)-2.75 F(currently)2.75 E
(speci\214ed, TFRC should not be used for v)100.8 205.8 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)100.8 218.8 Q 5.5
(.A)-1.221 G 2.75(tt)-5.5 G(he v)-2.75 E(ery least, safe operation with)
-.165 E -.275(va)100.8 231.8 S(lues of n greater than 8 w).275 E
(ould require a slight change to TFRC')-.11 E 2.75(sm)-.605 G
(echanisms to include)-2.75 E 2.75(am)100.8 244.8 S(ore se)-2.75 E -.165
(ve)-.275 G(re response to tw).165 E 2.75(oo)-.11 G 2.75(rm)-2.75 G
(ore round-trip times with hea)-2.75 E(vy pack)-.22 E(et loss.)-.11 E
(When calculating the a)100.8 261.4 Q -.165(ve)-.22 G(rage loss interv)
.165 E(al we need to decide whether to include the)-.275 E(interv)100.8
274.4 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)100.8 287.4 Q
-.165(ve)-.22 G(rage loss interv).165 E(al.)-.275 E
(Let the most recent loss interv)100.8 304 Q
(als be I_0 to I_k, where I_0 is the interv)-.275 E(al since the most)
-.275 E(recent loss e)100.8 317 Q -.165(ve)-.275 G 2.75(nt. If).165 F
(there ha)2.75 E .33 -.165(ve b)-.22 H(een at least n loss interv).165 E
(als, then k is set to n; otherwise k)-.275 E
(is the maximum number of loss interv)100.8 330 Q(als seen so f)-.275 E
(ar)-.11 E 5.5(.W)-.605 G 2.75(ec)-6.38 G(alculate the a)-2.75 E -.165
(ve)-.22 G(rage loss interv).165 E(al)-.275 E(I_mean is:)100.8 343 Q
(I_tot0 = 0;)136.8 362 Q(I_tot1 = 0;)136.8 375 Q(W_tot = 0;)136.8 388 Q
(for \(i = 0 to k-1\) {)136.8 401 Q(I_tot0 = I_tot0 + \(I_i * w_i\);)
142.3 414 Q(W_tot = W_tot + w_i;)142.3 427 Q(})136.8 440 Q
(for \(i = 1 to k\) {)136.8 453 Q
(I_tot1 = I_tot1 + \(I_i * w_\(i-1\)\);)142.3 466 Q(})136.8 479 Q
(I_tot = max\(I_tot0, I_tot1\);)136.8 492 Q(I_mean = I_tot/W_tot;)136.8
505 Q(The loss e)100.8 524 Q -.165(ve)-.275 G(nt rate, p is simply:).165
E 2.75(p=1/I)136.8 543 S(_mean;)-2.75 E/F1 11/Times-Bold@0 SF(5.5.)72
575 Q/F2 13/Times-Bold@0 SF(History Discounting)5.5 E F0
(As described in Section 5.4, the most recent loss interv)100.8 591.6 Q
(al is only assigned 1/\(0.75*n\) of the)-.275 E
(total weight in calculating the a)100.8 604.6 Q -.165(ve)-.22 G
(rage loss interv).165 E(al, re)-.275 E -.055(ga)-.165 G
(rdless of the size of the most recent).055 E(loss interv)100.8 617.6 Q
2.75(al. This)-.275 F
(section describes an optional history discounting mechanism, discussed)
2.75 E(further in [FHPW00a] and [W00], that allo)100.8 630.6 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,)-2.75 E(concentrating more of the relati)100.8
643.6 Q .33 -.165(ve w)-.275 H(eight on the most recent loss interv).165
E(al, when the most)-.275 E(recent loss interv)100.8 656.6 Q
(al is more than twice as lar)-.275 E(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
673.2 T(arry out history discounting, we associate a discount f).88 E
(actor DF_i with each loss)-.11 E(interv)100.8 686.2 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(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 15])-.165 E 0 Cg
EP
%%Page: 16 16
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(array maintains the cumulati)
100.8 85 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 98 Q(ginning, the v)-.165 E
(alues of DF_i in the discount array are initialized to 1:)-.275 E
(for \(i = 0 to n\) {)136.8 117 Q(DF_i = 1;)142.3 130 Q(})136.8 143 Q
(History discounting also uses a general discount f)100.8 162 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.)100.8 175 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)100.8 188 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)100.8 201 Q -.165
(ve)-.165 G 2.75(rt).165 G(ime.)-2.75 E
(As described in Section 5.4 the a)100.8 217.6 Q -.165(ve)-.22 G
(rage loss interv).165 E(al is calculated using the n pre)-.275 E
(vious loss)-.275 E(interv)100.8 230.6 Q
(als I_1, ..., I_n, and the interv)-.275 E
(al I_0 that represents the number of pack)-.275 E(ets recei)-.11 E
-.165(ve)-.275 G(d).165 E(since the last loss e)100.8 243.6 Q -.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 discount)-.275 E -.11(fa)100.8
256.6 S(ctors is a simple modi\214cation of the procedure in Section 5.\
4, as follo).11 E(ws:)-.275 E(I_tot0 = I_0 * w_0)136.8 275.6 Q
(I_tot1 = 0;)136.8 288.6 Q(W_tot0 = w_0)136.8 301.6 Q(W_tot1 = 0;)136.8
314.6 Q(for \(i = 1 to n-1\) {)136.8 327.6 Q
(I_tot0 = I_tot0 + \(I_i * w_i * DF_i * DF\);)142.3 340.6 Q
(W_tot0 = W_tot0 + w_i * DF_i * DF;)142.3 353.6 Q(})136.8 366.6 Q
(for \(i = 1 to n\) {)136.8 379.6 Q
(I_tot1 = I_tot1 + \(I_i * w_\(i-1\) * DF_i\);)142.3 392.6 Q
(W_tot1 = W_tot1 + w_\(i-1\) * DF_i;)142.3 405.6 Q(})136.8 418.6 Q 2.75
(p=m)136.8 431.6 S(in\(W_tot0/I_tot0, W_tot1/I_tot1\);)-2.75 E
(The general discounting f)100.8 450.6 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)100.8 463.6 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;)136.8 482.6 Q(W_tot = 0;)136.8 495.6 Q(for \(i = 1 to n\) {)
136.8 508.6 Q(W_tot = W_tot + w_\(i-1\) * DF_i;)142.3 521.6 Q
(I_tot = I_tot + \(I_i * w_\(i-1\) * DF_i\);)142.3 534.6 Q(})136.8 547.6
Q(I_mean = I_tot / W_tot;)136.8 560.6 Q(This weighted a)100.8 579.6 Q
-.165(ve)-.22 G(rage I_mean is compared to I_0, the number of pack).165
E(ets recei)-.11 E -.165(ve)-.275 G 2.75(ds).165 G(ince)-2.75 E
(the last loss e)100.8 592.6 Q -.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 interv)-2.75 E(al is)-.275 E(considerably lar)100.8 605.6 Q
(ger than the old ones, and the general discount f)-.198 E
(actor DF is updated to)-.11 E(decrease the relati)100.8 618.6 Q .33
-.165(ve w)-.275 H(eight on the older interv).165 E(als, as follo)-.275
E(ws:)-.275 E(Handle)72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E
(ye/W)-.055 E 189.593(idmer Section)-.44 F 2.75(5.5. [P)2.75 F(age 16])
-.165 E 0 Cg EP
%%Page: 17 17
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(if \(I_0 > 2 * I_mean\) {)
136.8 85 Q(DF = 2 * I_mean/I_0;)142.3 98 Q(if \(DF < THRESHOLD\))142.3
111 Q(DF = THRESHOLD;)147.8 124 Q 2.75(}e)136.8 137 S(lse)-2.75 E
(DF = 1;)142.3 150 Q 2.75(An)100.8 169 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)100.8 182 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)100.8 195 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.)100.8 208 Q
(When a ne)100.8 224.6 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)100.8 237.6 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)100.8 250.6 Q 5.5(.B)
-.715 G(ecause DF_i carries the discount f)-5.5 E(actor)-.11 E
(associated with loss interv)100.8 263.6 Q
(al I_i, the DF_i array has to be shifted as well. This is done as)-.275
E(follo)100.8 276.6 Q(ws:)-.275 E(for \(i = 1 to n\) {)136.8 295.6 Q
(DF_i = DF * DF_i;)142.3 308.6 Q(})136.8 321.6 Q
(for \(i = n-1 to 0 step -1\) {)136.8 334.6 Q(DF_\(i+1\) = DF_i;)142.3
347.6 Q(})136.8 360.6 Q(I_0 = 1;)136.8 373.6 Q(DF_0 = 1;)136.8 386.6 Q
(DF = 1;)136.8 399.6 Q(This completes the description of the optional h\
istory discounting mechanism. W)100.8 422.2 Q(e)-.88 E(emphasize that t\
his is an optional mechanism whose sole purpose is to allo)100.8 435.2 Q
2.75(wT)-.275 G(FRC to)-2.75 E(response some)100.8 448.2 Q(what more qu\
ickly to the sudden absence of congestion, as represented by a)-.275 E
(long current loss interv)100.8 461.2 Q(al.)-.275 E/F1 11/Times-Bold@0
SF(6.)72 500.2 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)100.8 516.8 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)100.8 529.8 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)100.8 542.8 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)100.8 555.8
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)100.8 568.8 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)100.8 581.8
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 \(man)100.8 598.4 Q
2.75(yp)-.165 G(ack)-2.75 E(ets per R)-.11 E(TT\) there may be some)-.66
E(adv)100.8 611.4 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)100.8 624.4 S
(ster response to changing R).11 E
(TT measurements, and more resilience to feedback pack)-.66 E(et)-.11 E
2.75(loss. Ho)100.8 637.4 R(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t)
.165 H(here is little g).44 E(ain from sending a lar)-.055 E
(ge number of feedback messages per)-.198 E -.66(RT)100.8 650.4 S -.814
(T.).66 G(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(6. [P)2.75 F(age 17])-.165 E 0
Cg EP
%%Page: 18 18
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E/F1 11/Times-Bold@0 SF(6.1.)72
85 Q/F2 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)100.8 101.6 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
118.2 R(the pack)2.75 E(et to the pack)-.11 E(et history)-.11 E(.)-.715
E 16.887(2\) Let)100.8 134.8 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
147.8 Q 16.887(3\) If)100.8 164.4 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
177.4 Q(If p <= p_pre)129.6 194 Q 2.75(vn)-.275 G 2.75(oa)-2.75 G
(ction need be performed.)-2.75 E(Ho)129.6 210.6 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)129.6 223.6 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
236.6 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)129.6 249.6 Q(fects of such an optimization are normally e)
-.275 E(xpected to be small.)-.165 E F1(6.2.)72 288.6 Q F2
(Expiration of feedback timer)5.5 E F0
(When the feedback timer at the recei)100.8 305.2 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)100.8 318.2 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 last feedback w)-2.75 E(as sent.)-.11 E
(Let the maximum sequence number of a pack)100.8 334.8 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)100.8 347.8 Q
(TT measurement included in pack)-.66 E(et S_m be R_m. If data pack)-.11
E(ets ha)-.11 E .33 -.165(ve b)-.22 H(een).165 E(recei)100.8 360.8 Q
-.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 2.75(rp).165 G
(erforms the follo)-2.75 E(wing steps:)-.275 E 16.887(1\) Calculate)
100.8 377.4 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)100.8 394 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)129.6 407 Q(vious R_m seconds.)-.275 E 16.887
(3\) Prepare)100.8 423.6 R(and send a feedback pack)2.75 E
(et containing the information described in Section)-.11 E(3.2.2)129.6
436.6 Q 16.887(4\) Restart)100.8 453.2 R(the feedback timer to e)2.75 E
(xpire after R_m seconds.)-.165 E(Note that rule 2\) abo)100.8 469.8 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)100.8 482.8 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
495.8 Q -.165(ve)-.275 G(nt rate, not from a limit imposed by the).165 E
(measured recei)100.8 508.8 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)100.8 525.4 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)100.8 538.4 Q
(xpire after R_m seconds.)-.165 E F1(6.3.)72 581 Q F2(Recei)5.5 E -.13
(ve)-.13 G 3.25(ri).13 G(nitialization)-3.25 E F0(The recei)100.8 597.6
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)100.8 610.6 Q(et be i.)-.11 E
(When the \214rst pack)100.8 627.2 Q(et is recei)-.11 E -.165(ve)-.275 G
(d:).165 E 24.95<8353>100.8 643.8 S(et p=0)-24.95 E 24.95<8353>100.8
660.4 S 2.75(et X_recv)-24.95 F 2.75(=0)2.75 G(.)-2.75 E 24.95<8350>
100.8 677 S(repare and send a feedback pack)-24.95 E(et.)-.11 E(Handle)
72 733 Q(y/Flo)-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 189.593
(idmer Section)-.44 F 2.75(6.3. [P)2.75 F(age 18])-.165 E 0 Cg EP
%%Page: 19 19
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 24.95<8353>100.8 85 S
(et the feedback timer to e)-24.95 E(xpire after R_i seconds.)-.165 E
(If the \214rst data pack)100.8 101.6 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)
100.8 114.6 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.)100.8 127.6 Q
(If the sender is using a coarse-grained timestamp that increments e)
100.8 144.2 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
157.2 Q(wing procedure from RFC 4342)-.275 E
(is used to determine when to send feedback messages.)100.8 170.2 Q
24.95<8357>100.8 186.8 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)129.6 199.8 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)129.6 212.8 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)129.6 225.8 Q(as sent.)-.11 E
24.95<8349>100.8 242.4 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)129.6 255.4 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. \("Greater")-.11 F
(and "greatest" are measured in circular windo)129.6 268.4 Q 2.75(wc)
-.275 G(ounter space.\))-2.75 E/F1 11/Times-Bold@0 SF 2.75
(6.3.1. Initializing)72 307.4 R
(the Loss History after the First Loss Ev)2.75 E(ent)-.11 E F0
(The number of pack)100.8 324 Q
(ets until the \214rst loss can not be used to compute the sending rate)
-.11 E(directly)100.8 337 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 sending rate when the loss occurred.)100.8 350 Q
(TFRC approximates this tar)100.8 363 Q(get rate by X_recv)-.198 E 2.75
(,t)-.715 G(he recei)-2.75 E .33 -.165(ve r)-.275 H(ate o).165 E -.165
(ve)-.165 G 2.75(rt).165 G(he most recent round-)-2.75 E(trip time.)
100.8 376 Q(After the \214rst loss, instead of initializing the \214rst\
 loss interv)5.5 E(al to the number of)-.275 E(pack)100.8 389 Q
(ets sent until the \214rst loss, the TFRC recei)-.11 E -.165(ve)-.275 G
2.75(rc).165 G(alculates the loss interv)-2.75 E(al that w)-.275 E
(ould be)-.11 E(required to produce the data rate X_recv)100.8 402 Q
2.75(,a)-.715 G(nd uses this synthetic loss interv)-2.75 E
(al to seed the)-.275 E(loss history mechanism.)100.8 415 Q
(TFRC does this by \214nding some v)100.8 441 Q
(alue p for which the throughput equation in Section 3.1)-.275 E(gi)
100.8 454 Q -.165(ve)-.275 G 2.75(sas).165 G
(ending rate within 5% of X_recv)-2.75 E 2.75(,g)-.715 G -2.365 -.275
(iv e)-2.75 H 2.75(nt).275 G(he round-trip time R, and the \214rst loss)
-2.75 E(interv)100.8 467 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)100.8 480 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 meaure the)-2.75 E(recei)100.8 493 Q
.33 -.165(ve r)-.275 H(ate in pack).165 E(ets per second instead of byt\
es per second for this purpose, and use the)-.11 E
(throughput equation for X_pps.)100.8 506 Q
(\(The 5% tolerance is introduced simply because the)5.5 E
(throughput equation is dif)100.8 519 Q(\214cult to in)-.275 E -.165(ve)
-.44 G(rt, and we w).165 E(ant to reduce the costs of calculating p)-.11
E(numerically)100.8 532 Q(.\))-.715 E(If the \214rst arri)100.8 558 Q
(ving data pack)-.275 E(et is lost or ECN-mark)-.11 E
(ed, then the loss e)-.11 E -.165(ve)-.275 G(nt consists of the).165 E
(*** SALL)100.8 571 Q 2.75(Y*)-1.1 G(**)-2.75 E F1(7.)72 610 Q/F2 14
/Times-Bold@0 SF(Sender)5.5 E(-based V)-.518 E(ariants)-1.288 E F0(It w)
100.8 626.6 Q(ould be possible to implement a sender)-.11 E(-based v)
-.22 E(ariant of TFRC, where the recei)-.275 E -.165(ve)-.275 G 2.75(ru)
.165 G(ses)-2.75 E(reliable deli)100.8 639.6 Q -.165(ve)-.275 G
(ry to send information about pack).165 E(et losses to the sender)-.11 E
2.75(,a)-.44 G(nd the sender)-2.75 E(computes the pack)100.8 652.6 Q
(et loss rate and the acceptable transmit rate.)-.11 E(Ho)5.5 E(we)-.275
E -.165(ve)-.275 G .88 -.44(r, w).165 H 2.75(ed).44 G 2.75(on)-2.75 G
(ot specify)-2.75 E(the details of a sender)100.8 665.6 Q(-based v)-.22
E(ariant in this document.)-.275 E(The main adv)100.8 691.6 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(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(7. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(not ha)100.8 85 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 98 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)100.8 111
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.)100.8 124 Q(In contrast, the recei)100.8 150 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)100.8 163 Q
(ets, and therefore does not require the reliable deli)-.11 E -.165(ve)
-.275 G(ry of feedback).165 E(pack)100.8 176 Q 2.75(ets. It)-.11 F(is a\
lso better suited for applications such as streaming media from web ser\
v)2.75 E(ers,)-.165 E(where it is typically desirable to of)100.8 189 Q
(\215oad w)-.275 E(ork from the serv)-.11 E(er to the client as much as)
-.165 E(possible.)100.8 202 Q(The sender)100.8 228 Q(-based and recei)
-.22 E -.165(ve)-.275 G -.22(r-).165 G(based v).22 E(ariants also ha)
-.275 E .33 -.165(ve d)-.22 H(if).165 E(ferent properties in terms of)
-.275 E 2.75(upgrades. F)100.8 241 R(or e)-.165 E
(xample, for changes in the procedure for calculating the pack)-.165 E
(et loss rate,)-.11 E(the sender w)100.8 254 Q(ould ha)-.11 E .33 -.165
(ve t)-.22 H 2.75(ob).165 G 2.75(eu)-2.75 G(pgraded in the sender)-2.75
E(-based v)-.22 E(ariant, and the recei)-.275 E -.165(ve)-.275 G 2.75
(rw).165 G(ould)-2.86 E(ha)100.8 267 Q .33 -.165(ve t)-.22 H 2.75(ob)
.165 G 2.75(eu)-2.75 G(pgraded in the recei)-2.75 E -.165(ve)-.275 G
-.22(r-).165 G(based v).22 E(ariant.)-.275 E/F1 11/Times-Bold@0 SF(8.)72
306 Q/F2 14/Times-Bold@0 SF(Implementation Issues)5.5 E F0(This documen\
t has speci\214ed the TFRC congestion control mechanism, for use by)
100.8 322.6 Q(applications and transport protocols.)100.8 335.6 Q
(This section mentions brie\215y some of the fe)5.5 E(w)-.275 E
(implementation issues.)100.8 348.6 Q -.165(Fo)100.8 374.6 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)100.8 387.6 Q(ws:)-.275 E(s)158.8 406.6 Q
-2.75 2.75(X= -)136.8 419.6 T(-------)-2.75 E 2.75(R*f)150.55 432.6 S
(\(p\))-2.75 E(for)100.8 451.6 Q(f\(p\) =)136.8 470.6 Q
(sqrt\(2*p/3\) + \(12*sqrt\(3*p/8\) * p * \(1+32*p^2\)\).)5.5 E 2.75(At)
100.8 489.6 S(able lookup could be used for the function f\(p\).)-2.75 E
(Man)100.8 515.6 Q 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he multiplications \(e.g., q and 1-q for the round-trip time a)-2.75 E
-.165(ve)-.22 G(rage, a f).165 E(actor of 4)-.11 E
(for the timeout interv)100.8 528.6 Q(al\) 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.)100.8 541.6 Q 1.76 -.88(We n)
100.8 567.6 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.)100.8 580.6 Q
(The calculation of the a)100.8 606.6 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:)100.8 619.6 Q
(1.0, 1.0, 1.0, 1.0, 0.8, 0.6, 0.4, 0.2.)129.6 636.2 Q -.44(Wi)100.8
652.8 S(th a minor loss of smoothness, it w).44 E
(ould be possible to use weights that were po)-.11 E(wers of)-.275 E(tw)
100.8 665.8 Q 2.75(oo)-.11 G 2.75(rs)-2.75 G(ums of po)-2.75 E
(wers of tw)-.275 E(o, e.g.,)-.11 E
(1.0, 1.0, 1.0, 1.0, 0.75, 0.5, 0.25, 0.25.)129.6 682.4 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 20])-.165 E 0 Cg EP
%%Page: 21 21
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(The optional history discount\
ing mechanism described in Section 5.5 is used in the)100.8 85 Q
(calculation of the a)100.8 98 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)100.8 111 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)100.8 124 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.)72 163 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)100.8 179.6 Q(ws:)-.275 E 24.95
<8343>100.8 196.2 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)129.6 209.2
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)129.6 222.2 Q(ets per round trip time, follo)-.11 E
(wing RFC 3390.)-.275 E -.165(Fo)129.6 248.2 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)
129.6 261.2 Q 2.75(wt)-.275 G(he initial sending)-2.75 E 2.75(rate. In)
129.6 274.2 R(addition, when the sender has been data-limited and the s\
ender is reducing)2.75 E(the allo)129.6 287.2 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)
129.6 300.2 Q(wed transmit rate to less than the initial sending rate.)
-.275 E 2.75(Al)129.6 326.2 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)129.6 339.2 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)129.6 352.2 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)129.6 365.2 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)129.6 378.2
Q(et.)-.11 E 24.95<8352>100.8 394.8 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)129.6 407.8 Q(o* round-trip times without recei)-.11 E
(ving a feedback report, or after *four*)-.275 E(round-trip times.)129.6
420.8 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)
129.6 433.8 Q(ving a feedback report [RFC3448Err].)-.275 E 24.95<8353>
100.8 450.4 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)129.6 463.4 S 2.75(,s)-2.75 G
(tep \(2\) applies instead of step \(1\) [RFC3448Err].)-2.75 E 24.95
<8341>100.8 480 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; }')129.6 493 Q 2.75('[)
-.814 G(RFC3448Err].)-2.75 E 24.95<8353>100.8 509.6 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)129.6 522.6 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>100.8 539.2 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)
129.6 552.2 Q -.165(ve)-.22 G(rage se).165 E(gment size.)-.165 E F1(10.)
72 578.2 Q F2(Security Considerations)5.5 E F0
(TFRC is not a transport protocol in its o)100.8 594.8 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
607.8 Q(Therefore security)5.5 E
(primarily needs to be considered in the conte)100.8 620.8 Q
(xt of a speci\214c transport protocol and its)-.165 E
(authentication mechanisms.)100.8 633.8 Q
(Congestion control mechanisms can potentially be e)100.8 650.4 Q
(xploited to create denial of service.)-.165 E
(This may occur through spoofed feedback.)100.8 663.4 Q(Thus an)5.5 E
2.75(yt)-.165 G(ransport protocol that uses TFRC)-2.75 E(should tak)
100.8 676.4 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)100.8 689.4 Q .33 -.165(ve t)-.275 H
(his will ho).165 E(we)-.275 E -.165(ve)-.275 G 2.75(rd).165 G
(epend on the transport protocol itself.)-2.75 E(Handle)72 733 Q(y/Flo)
-.165 E(yd/P)-.11 E(adh)-.165 E(ye/W)-.055 E 192.343(idmer Section)-.44
F 2.75(10. [P)2.75 F(age 21])-.165 E 0 Cg EP
%%Page: 22 22
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(In addition, congestion contr\
ol mechanisms may potentially be manipulated by a greedy)100.8 85 Q
(recei)100.8 98 Q -.165(ve)-.275 G 2.75(rt).165 G(hat wishes to recei)
-2.75 E .33 -.165(ve m)-.275 H(ore than its f).165 E(air share of netw)
-.11 E(ork bandwidth.)-.11 E 2.75(Ar)5.5 G(ecei)-2.75 E -.165(ve)-.275 G
(r).165 E(might do this by claiming to ha)100.8 111 Q .33 -.165(ve r)
-.22 H(ecei).165 E -.165(ve)-.275 G 2.75(dp).165 G(ack)-2.75 E
(ets that in f)-.11 E(act were lost due to congestion.)-.11 E
(Possible defenses ag)100.8 124 Q(ainst such a recei)-.055 E -.165(ve)
-.275 G 2.75(rw).165 G(ould normally include some form of nonce that)
-2.86 E(the recei)100.8 137 Q -.165(ve)-.275 G 2.75(rm).165 G
(ust feed back to the sender to pro)-2.75 E .33 -.165(ve r)-.165 H 2.75
(eceipt. Ho).165 F(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H
(he details of such a).44 E(nonce w)100.8 150 Q(ould depend on the tran\
sport protocol, and in particular on whether the transport)-.11 E
(protocol is reliable or unreliable.)100.8 163 Q 1.76 -.88(We e)100.8
179.6 T(xpect that protocols incorporating ECN with TFRC will also w)
.715 E(ant to incorporate)-.11 E(feedback from the recei)100.8 192.6 Q
-.165(ve)-.275 G 2.75(rt).165 G 2.75(ot)-2.75 G
(he sender using the ECN nonce [RFC3540].)-2.75 E(The ECN)8.25 E(nonce \
is a modi\214cation to ECN that protects the sender from the accidental\
 or malicious)100.8 205.6 Q(concealment of mark)100.8 218.6 Q(ed pack)
-.11 E 2.75(ets. Ag)-.11 F(ain, the details of such a nonce w)-.055 E
(ould depend on the)-.11 E
(transport protocol, and are not addressed in this document.)100.8 231.6
Q/F1 11/Times-Bold@0 SF(11.)72 270.6 Q/F2 14/Times-Bold@0 SF(IAN)5.5 E
3.5(AC)-.28 G(onsiderations)-3.5 E F0(There are no IAN)100.8 287.2 Q
2.75(Aa)-.385 G(ctions required for this document.)-2.75 E F1(12.)72
326.2 Q F2(Ackno)5.5 E(wledgments)-.14 E F0 1.76 -.88(We w)100.8 342.8 T
(ould lik).77 E 2.75(et)-.11 G 2.75(oa)-2.75 G(ckno)-2.75 E
(wledge feedback and discussions on equation-based congestion)-.275 E(c\
ontrol with a wide range of people, including members of the Reliable M\
ulticast Research)100.8 355.8 Q(Group, the Reliable Multicast T)100.8
368.8 Q(ransport W)-.385 E(orking Group, and the End-to-End Research)
-.88 E 5.5(Group. W)100.8 381.8 R 2.75(ew)-.88 G(ould lik)-2.86 E 2.75
(et)-.11 G 2.75(ot)-2.75 G(hank Dado Colussi, W)-2.75 E(im Heirman, K)
-.44 E(en Lofgren, Mik)-.275 E 2.75(eL)-.11 G(uby)-2.75 E(,)-.715 E
(Michele R., Vladica Stanisic, Randall Ste)100.8 394.8 Q -.11(wa)-.275 G
(rt, Eduardo Urzaiz, Shushan W).11 E(en, and W)-.88 E(endy)-.88 E
(Lee \(lhh@zsu.edu.cn\) for feedback on earlier v)100.8 407.8 Q
(ersions of this document, and to thank Mark)-.165 E(Allman for his e)
100.8 420.8 Q(xtensi)-.165 E .33 -.165(ve f)-.275 H
(eedback from using the document to produce a w).165 E(orking)-.11 E
(implementation.)100.8 433.8 Q F1(13.)72 459.8 Q F2(Normati)5.5 E .28
-.14(ve R)-.14 H(efer).14 E(ences)-.252 E F1(14.)72 485.8 Q F2(Inf)5.5 E
(ormational Refer)-.35 E(ences)-.252 E F0 40.634([BRS99] Balakrishnan,)
113.1 502.4 R(H., Rahul, H., and Seshan, S., "An Inte)2.75 E(grated)
-.165 E(Congestion Management Architecture for Internet Hosts," Proc.)
195.6 515.4 Q -.44(AC)195.6 528.4 S 2.75(MS).44 G
(IGCOMM, Cambridge, MA, September 1999.)-2.75 E 30.866([FHPW00] S.)113.1
545 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
558 Q(SIGCOMM 2000.)195.6 571 Q 25.982([FHPW00a] S.)113.1 587.6 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)195.6 600.6
Q(ersion",)-1.221 E(ICSI tech report TR-00-03, March 2000.)195.6 613.6 Q
34.529([PFTK98] P)113.1 630.2 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)195.6 643.2 Q
(alidation",)-1.221 E(Proc A)195.6 656.2 Q(CM SIGCOMM 1998.)-.44 E
29.634([RFC2119] S.)113.1 672.8 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)195.6 685.8 Q -.165
(ve)-.275 G(ls, RFC 2119.).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(14. [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 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E 29.634([RFC2140] J.)113.1 85 R
-.88(To)2.75 G
(uch, "TCP Control Block Interdependence", RFC 2140, April).88 E(1997.)
195.6 98 Q 29.634([RFC2988] V)113.1 114.6 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)195.6 127.6 S(mer", RFC 2988, No).385 E
-.165(ve)-.165 G(mber 2000.).165 E 29.634([RFC3168] K.)113.1 144.2 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.)195.6 157.2 Q
29.634([RFC3390] Allman,)113.1 173.8 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)195.6 186.8 S(ndo).44 E(w", RFC 3390, October 2002.)-.275 E
15.587([RFC3448Err] RFC)113.1 203.4 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)113.1 220 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)195.6 233 Q 29.634
([RFC4340] K)113.1 249.6 R(ohler)-.385 E 2.75(,E)-.44 G(., Handle)-2.75
E 1.43 -.715(y, M)-.165 H(., and S. Flo).715 E(yd, "Datagram Congestion)
-.11 E(Control Protocol \(DCCP\)", RFC 4340, March 2006.)195.6 262.6 Q
29.634([RFC4342] Flo)113.1 279.2 R(yd, S., K)-.11 E(ohler)-.385 E 2.75
(,E)-.44 G(., and J. P)-2.75 E(adh)-.165 E(ye, "Pro\214le for Datagram)
-.055 E
(Congestion Control Protocol \(DCCP\) Congestion Control ID 3: TCP-)
195.6 292.2 Q(Friendly Rate Control \(TFRC\)", RFC 4342, March 2006.)
195.6 305.2 Q 51.04([W00] W)113.1 321.8 R(idmer)-.44 E 2.75(,J)-.44 G
(., "Equation-Based Congestion Control", Diploma Thesis,)-2.75 E(Uni)
195.6 334.8 Q -.165(ve)-.275 G(rsity of Mannheim, February 2000.).165 E
(URL)5.5 E("http://www)195.6 347.8 Q(.icir)-.715 E(.or)-.605 E
(g/tfrc/".)-.198 E/F1 11/Times-Bold@0 SF(15.)72 386.8 Q/F2 14
/Times-Bold@0 SF -.7(Au)5.5 G(thors' Addr).7 E(esses)-.252 E F0(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 23])-.165 E 0 Cg EP
%%Page: 24 24
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(Mark Handle)136.8 85 Q -.715
(y,)-.165 G(Department of Computer Science)136.8 98 Q(Uni)136.8 111 Q
-.165(ve)-.275 G(rsity Colle).165 E(ge London)-.165 E(Go)136.8 124 Q
(wer Street)-.275 E(London WC1E 6BT)136.8 137 Q(UK)136.8 150 Q
(EMail: M.Handle)136.8 163 Q(y@cs.ucl.ac.uk)-.165 E(Sally Flo)136.8 189
Q(yd)-.11 E(ICIR/ICSI)136.8 202 Q(1947 Center St, Suite 600)136.8 215 Q
(Berk)136.8 228 Q(ele)-.11 E 1.43 -.715(y, C)-.165 H 2.75(A9).715 G
(4708)-2.75 E<8d6f>136.8 241 Q(yd@icir)-.11 E(.or)-.605 E(g)-.198 E
(Jitendra P)136.8 267 Q(adh)-.165 E(ye)-.055 E(Microsoft Research)136.8
280 Q(padh)136.8 293 Q(ye@microsoft.com)-.055 E .918<4aa8>136.8 332 S
(or)-5.499 E 2.75(gW)-.198 G(idmer)-3.19 E
(Lehrstuhl Praktische Informatik IV)136.8 345 Q(Uni)136.8 358 Q -.165
(ve)-.275 G -1.832(rsit \250).165 F(at Mannheim)-4.581 E 2.75(L1)136.8
371 S(5, 16 - Room 415)-2.75 E(D-68131 Mannheim)136.8 384 Q(German)136.8
397 Q(y)-.165 E(widmer@informatik.uni-mannheim.de)136.8 410 Q/F1 14
/Times-Bold@0 SF(Full Copyright Statement)72 455 Q F0(Cop)100.8 471.6 Q
(yright \(C\) The Internet Society 2006.)-.11 E
(This document is subject to the rights, licenses)5.5 E
(and restrictions contained in BCP 78, and e)100.8 484.6 Q
(xcept as set forth therein, the authors retain all)-.165 E
(their rights.)100.8 497.6 Q
(This document and the information contained herein are pro)100.8 514.2
Q(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)100.8 527.2 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 AND THE INTERNET)100.8
540.2 Q(ENGINEERING T)100.8 553.2 Q(ASK FORCE DISCLAIM ALL W)-1.023 E
(ARRANTIES, EXPRESS OR)-1.32 E(IMPLIED, INCLUDING B)100.8 566.2 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 THA)-1.32 E 2.75(TT)-1.221 G(HE USE)-2.75 E(OF THE INFORMA)
100.8 579.2 Q(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G
(NFRINGE ANY RIGHTS OR ANY)-2.75 E(IMPLIED W)100.8 592.2 Q
(ARRANTIES OF MERCHANT)-1.32 E(ABILITY OR FITNESS FOR A)-1.023 E -1.012
(PA)100.8 605.2 S -.66(RT)1.012 G(ICULAR PURPOSE.).66 E F1
(Intellectual Pr)72 631.2 Q(operty)-.252 E F0(The IETF tak)100.8 647.8 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 claimed to pertain to the \
implementation or use of the)100.8 660.8 Q
(technology described in this document or the e)100.8 673.8 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)100.8 686.8 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(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 24])-.165 E 0 Cg EP
%%Page: 25 25
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 77.081(-DRAFT Expires:)-1.012 F
(April 2007)2.75 E(October 2006)112.127 E(ef)100.8 85 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 98 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)100.8
114.6 Q 2.75(ya)-.165 G(ssurances of licenses to)-2.75 E(be made a)100.8
127.6 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)100.8 140.6 Q(speci\214cati\
on can be obtained from the IETF on-line IPR repository at)100.8 153.6 Q
(http://www)100.8 166.6 Q(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E
(The IETF in)100.8 183.2 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
196.2 Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)100.8 209.2 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
100.8 222.2 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 25])-.165 E 0 Cg EP
%%Trailer
end
%%EOF