One document matched: draft-ietf-tcpm-ecnsyn-02.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.18.1
%%CreationDate: Sat Jun 30 21:35:24 2007
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.18 1
%%Pages: 20
%%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{10 -52 translate
/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 98 Q
(A. Kuzmanovic)167.999 E 287.999(INTERNET-DRAFT A.)72 111 R(Mondal)6 E
(Intended status: Proposed Standard)72 124 Q(Northwestern University)
89.999 E(Expires: 30 December 2007)72 137 Q(S. Floyd)233.999 E(ICIR)
479.999 150 Q(K.K. Ramakrishnan)401.999 163 Q(AT&T)479.999 176 Q
(30 June 2007)431.999 189 Q/F1 14/Times-Bold@0 SF
(Adding Explicit Congestion Noti\214cation \(ECN\) Capability to TCP')
86.218 227 Q(s)-.518 E(SYN/A)232.336 240 Q(CK P)-.77 E(ack)-.14 E(ets)
-.14 E(draft-ietf-tcpm-ecnsyn-02.ps)203.048 253 Q/F2 11/Times-Bold@0 SF
(Status of this Memo)72 285 Q/F3 11/Times-Roman@0 SF
(By submitting this Internet-Draft, each author represents that an)93.6
301.6 Q 2.75(ya)-.165 G(pplicable patent or other)-2.75 E
(IPR claims of which he or she is a)93.6 314.6 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)93.6 327.6
Q -.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)93.6 353.6 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)93.6 366.6 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.)93.6 379.6 Q
(Internet-Drafts are draft documents v)93.6 405.6 Q
(alid for a maximum of six months and may be updated,)-.275 E
(replaced, or obsoleted by other documents at an)93.6 418.6 Q 2.75(yt)
-.165 G 2.75(ime. It)-2.75 F(is inappropriate to use Internet-)2.75 E
(Drafts as reference material or to cite them other than as "w)93.6
431.6 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)93.6
457.6 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)93.6 470.6 Q
(The list of Internet-Draft Shado)93.6 496.6 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)93.6 509.6 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)93.6 535.6 Q(xpire on December 2007.)-.165 E
F2(Copyright Notice)72 561.6 Q F3(Cop)93.6 578.2 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E F2(Abstract)72
604.2 Q F3(This draft speci\214es a modi\214cation to RFC 3168 to allo)
93.6 620.8 Q 2.75(wT)-.275 G(CP SYN/A)-2.75 E(CK pack)-.44 E
(ets to be ECN-)-.11 E 2.75(Capable. F)93.6 633.8 R(or TCP)-.165 E 2.75
(,R)-1.221 G
(FC 3168 only speci\214es setting an ECN-Capable codepoint on data)-2.75
E(pack)93.6 646.8 Q(ets, and not on SYN and SYN/A)-.11 E(CK pack)-.44 E
2.75(ets. Ho)-.11 F(we)-.275 E -.165(ve)-.275 G .88 -.44(r, b).165 H
(ecause of the high cost to the).44 E(TCP transfer of ha)93.6 659.8 Q
(ving a SYN/A)-.22 E(CK pack)-.44 E
(et dropped, with the resulting retransmit timeout,)-.11 E
(this document speci\214es the use of ECN for the SYN/A)93.6 672.8 Q
(CK pack)-.44 E(et itself, when sent in)-.11 E(response to a SYN pack)
93.6 685.8 Q(et with the tw)-.11 E 2.75(oE)-.11 G
(CN \215ags set in the TCP header)-2.75 E 2.75(,i)-.44 G(ndicating a)
-2.75 E(Flo)72 733 Q 366.891(yd [P)-.11 F(age 1])-.165 E 0 Cg EP
%%Page: 2 2
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(willingness to use ECN.)93.6 85 Q(Setting TCP SYN/A)5.5 E
(CK pack)-.44 E(ets as ECN-Capable can be of great)-.11 E
(bene\214t to the TCP connection, a)93.6 98 Q -.22(vo)-.22 G
(iding the se).22 E -.165(ve)-.275 G
(re penalty of a retransmit timeout for a).165 E
(connection that has not yet started placing a load on the netw)93.6 111
Q 2.75(ork. The)-.11 F(sender of the)2.75 E(SYN/A)93.6 124 Q(CK pack)
-.44 E(et must respond to a report of an ECN-mark)-.11 E(ed SYN/A)-.11 E
(CK pack)-.44 E(et by)-.11 E(reducing its initial congestion windo)93.6
137 Q 2.75(wf)-.275 G(rom tw)-2.75 E(o, three, or four se)-.11 E
(gments to one se)-.165 E(gment,)-.165 E
(thereby reducing the subsequent load from that connection on the netw)
93.6 150 Q(ork.)-.11 E/F1 11/Times-Bold@0 SF -1.012(Ta)72 176 S
(ble of Contents)1.012 E/F2 10/Times-Roman@0 SF(1. Con)93.6 201 Q -.15
(ve)-.4 G(ntions).15 E F0 -2.25(.......................................\
.......................................................................\
.......... ....)3.89 F F2(3)A(2. Introduction)93.6 213 Q F0 -2.25(.....\
.......................................................................\
............................................ ....)4.46 F F2(3)A
(3. Proposal)93.6 225 Q F0 -2.25(......................................\
.......................................................................\
................ ....)5.15 F F2(5)A(4. Discussion)93.6 237 Q F0 -2.25(.\
.......................................................................\
.................................................. ....)4.51 F F2(7)A
(5. Related W)93.6 249 Q(ork)-.8 E F0 -1.75(...........................\
.......................................................................\
.................... ..)4.38 F F2(10)A(6. Performance Ev)93.6 261 Q
(aluation)-.25 E F0 -1.75(.............................................\
.......................................................... ..)3.98 F F2
(11)A(6.1. The Costs and Bene\214t of Adding ECN-Capability)108.6 273 Q
F0 -1.75(......................................................... ..)
5.21 F F2(11)A(6.2. An Ev)108.6 285 Q(aluation of Dif)-.25 E
(ferent Responses to ECN-Mark)-.25 E(ed SYN/A)-.1 E(CK P)-.4 E(ack)-.15
E(ets)-.1 E F0 -1.75(...................... ..)2.75 F F2(12)A
(7. Security Considerations)93.6 297 Q F0 -1.75(.......................\
.......................................................................\
......... ..)4.82 F F2(13)A(8. Conclusions)93.6 309 Q F0 -1.75(........\
.......................................................................\
......................................... ..)4.45 F F2(14)A(9. Ackno)
93.6 321 Q(wledgements)-.25 E F0 -1.75(................................\
.......................................................................\
...... ..)4.97 F F2(14)A(A. Report on Simulations)93.6 333 Q F0 -1.75(.\
.......................................................................\
................................. ..)2.92 F F2(14)A
(A.1. Simulations with RED in P)108.6 345 Q(ack)-.15 E(et Mode)-.1 E F0
-1.75(.................................................................\
........ ..)2.84 F F2(15)A(A.2. Simulations with RED in Byte Mode)108.6
357 Q F0 -1.75(........................................................\
................... ..)4.86 F F2(16)A(Normati)93.6 369 Q .3 -.15(ve R)
-.25 H(eferences).15 E F0 -1.75(.......................................\
...................................................................... \
..)4.86 F F2(17)A(Informati)93.6 381 Q .3 -.15(ve R)-.25 H(eferences).15
E F0 -1.75(............................................................\
................................................ ..)3.17 F F2(17)A(IAN)
93.6 393 Q 2.5(AC)-.35 G(onsiderations)-2.5 E F0 -1.75(................\
.......................................................................\
....................... ..)4.26 F F2(18)A(Full Cop)93.6 405 Q
(yright Statement)-.1 E F0 -1.75(......................................\
................................................................... ..)
3.57 F F2(19)A(Intellectual Property)93.6 417 Q F0 -1.75(..............\
.......................................................................\
........................... ..)3.97 F F2(19)A F0(NO)93.6 446.6 Q(TE T)
-.44 E 2.75(OR)-.198 G(FC EDIT)-2.75 E(OR: PLEASE DELETE THIS NO)-.198 E
(TE UPON PUBLICA)-.44 E(TION.)-1.221 E
(Changes from draft-ietf-tcpm-ecnsyn-01:)93.6 472.6 Q 2.75(*C)93.6 498.6
S(hanges in response to feedback from Anil Ag)-2.75 E(arw)-.055 E(al.)
-.11 E 2.75(*A)93.6 524.6 S
(dded a look at the costs of adding ECN-Capability to)-2.75 E(SYN/A)99.1
537.6 Q(CKs in a highly-congested scenario.)-.44 E
(From feedback from Mark Allman and Janardhan Iyeng)99.1 550.6 Q(ar)
-.055 E(.)-.605 E 2.75(*A)93.6 576.6 S(dded a comparati)-2.75 E .33
-.165(ve e)-.275 H -.275(va)-.11 G(luation of tw).275 E 2.75(op)-.11 G
(ossible responses)-2.75 E(to an ECN-mark)99.1 589.6 Q(ed SYN/A)-.11 E
(CK pack)-.44 E 2.75(et. From)-.11 F(Mark Allman.)2.75 E
(Changes from draft-ietf-tcpm-ecnsyn-00:)93.6 615.6 Q 2.75(*O)93.6 641.6
S(nly updating the re)-2.75 E(vision number)-.275 E(.)-.605 E
(Changes from draft-ietf-twvsg-ecnsyn-00:)93.6 667.6 Q 2.75(*C)93.6
693.6 S(hanged name of draft to draft-ietf-tcpm-ecnsyn.)-2.75 E(Flo)72
733 Q 366.891(yd [P)-.11 F(age 2])-.165 E 0 Cg EP
%%Page: 3 3
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E 2.75(*A)93.6 85 S
(dded a discussion in Section 3 of "Response to)-2.75 E
(ECN-marking of SYN/A)99.1 98 Q(CK pack)-.44 E 2.75(ets". Based)-.11 F
(on)2.75 E(suggestions from Mark Allman.)99.1 111 Q 2.75(*A)93.6 137 S
(dded a discussion to the Conclusions about adding)-2.75 E
(ECN-capability to rele)99.1 150 Q -.275(va)-.275 G(nt set-up pack).275
E(ets in other)-.11 E 2.75(protocols. From)99.1 163 R 2.75(as)2.75 G
(uggestion from W)-2.75 E(esle)-.88 E 2.75(yE)-.165 G(ddy)-2.75 E(.)
-.715 E 2.75(*A)93.6 189 S(dded a description of SYN e)-2.75 E
(xchanges with SYN cookies.)-.165 E(From a suggestion from W)99.1 202 Q
(esle)-.88 E 2.75(yE)-.165 G(ddy)-2.75 E(.)-.715 E 2.75(*A)93.6 228 S
(dded a discussion of one-w)-2.75 E(ay data transfers, where the)-.11 E
(host sending the SYN/A)99.1 241 Q(CK pack)-.44 E(et sends no data pack)
-.11 E(ets.)-.11 E 2.75(*M)93.6 267 S
(inor editing, from feedback from Mark Allman and Janardhan)-2.75 E
(Iyeng)99.1 280 Q(ar)-.055 E(.)-.605 E 2.75(*F)93.6 306 S(uture w)-2.75
E(ork: a look at the costs of adding)-.11 E(ECN-Capability in a w)99.1
319 Q(orst-case scenario.)-.11 E
(From feedback from Mark Allman and Janardhan Iyeng)99.1 332 Q(ar)-.055
E(.)-.605 E 2.75(*F)93.6 358 S(uture w)-2.75 E(ork: a comparati)-.11 E
.33 -.165(ve e)-.275 H -.275(va)-.11 G(luation of tw).275 E(o)-.11 E
(possible responses to an ECN-mark)99.1 371 Q(ed SYN/A)-.11 E(CK pack)
-.44 E(et.)-.11 E(Changes from draft-kuzmano)93.6 397 Q
(vic-ecn-syn-00.txt:)-.165 E 2.75(*C)93.6 423 S
(hanged name of draft to draft-ietf-twvsg-ecnsyn.)-2.75 E(END OF NO)93.6
449 Q(TE T)-.44 E 2.75(OR)-.198 G(FC EDIT)-2.75 E(OR.)-.198 E/F1 11
/Times-Bold@0 SF(1.)72 475 Q/F2 14/Times-Bold@0 SF(Con)5.5 E -.14(ve)
-.56 G(ntions).14 E F0(The k)93.6 501 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)93.6
514 Q(T", "RECOMMENDED", "MA)-.44 E(Y", and "OPTION)-1.155 E
(AL" in this)-.385 E
(document are to be interpreted as described in [RFC 2119].)93.6 527 Q
F1(2.)72 553 Q F2(Intr)5.5 E(oduction)-.252 E F0(TCP')93.6 579 Q 2.75
(sc)-.605 G(ongestion control mechanism has primarily used pack)-2.75 E
(et loss as the congestion)-.11 E(indication, with pack)93.6 592 Q
(ets dropped when b)-.11 E(uf)-.22 E(fers o)-.275 E -.165(ve)-.165 G
(r\215o).165 E 4.18 -.715(w. W)-.275 H(ith such tail-drop mechanisms,)
.275 E(the pack)93.6 605 Q
(et delay can be high, as the queue at bottleneck routers can be f)-.11
E(airly lar)-.11 E 2.75(ge. Dropping)-.198 F(pack)93.6 618 Q
(ets only when the queue o)-.11 E -.165(ve)-.165 G(r\215o).165 E
(ws, and ha)-.275 E(ving TCP react only to such losses, results in:)-.22
E(1\) signi\214cantly higher pack)93.6 631 Q(et delay;)-.11 E
(2\) unnecessarily man)93.6 644 Q 2.75(yp)-.165 G(ack)-2.75 E
(et losses; and)-.11 E(3\) unf)93.6 657 Q
(airness due to synchronization ef)-.11 E(fects.)-.275 E
(The adoption of Acti)93.6 683 Q .33 -.165(ve Q)-.275 H
(ueue Management \(A).165 E(QM\) mechanisms allo)-.605 E
(ws better control of)-.275 E(bottleneck queues [RFC2309].)93.6 696 Q
(This use of A)5.5 E(QM has the follo)-.605 E
(wing potential bene\214ts:)-.275 E(Flo)72 733 Q 366.891(yd [P)-.11 F
(age 3])-.165 E 0 Cg EP
%%Page: 4 4
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E
(1\) better control of the queue, with reduced queueing delay;)93.6 85 Q
(2\) fe)93.6 98 Q(wer pack)-.275 E(et drops; and)-.11 E(3\) better f)
93.6 111 Q(airness because of fe)-.11 E(wer synchronization ef)-.275 E
(fects.)-.275 E -.44(Wi)93.6 137 S
(th the adoption of ECN, performance may be further impro).44 E -.165
(ve)-.165 G 2.75(d. When).165 F(the router detects)2.75 E
(congestion before b)93.6 150 Q(uf)-.22 E(fer o)-.275 E -.165(ve)-.165 G
(r\215o).165 E 1.43 -.715(w, t)-.275 H(he router can pro).715 E
(vide a congestion indication either by)-.165 E(dropping a pack)93.6 163
Q(et, or by setting the Congestion Experienced \(CE\) codepoint in the)
-.11 E(Explicit)5.5 E
(Congestion Noti\214cation \(ECN\) \214eld in the IP header [RFC3168].)
93.6 176 Q(The IETF has)5.5 E(standardized the use of the Congestion Ex\
perienced \(CE\) codepoint in the IP header for)93.6 189 Q
(routers to indicate congestion.)93.6 202 Q -.165(Fo)5.5 G 2.75(ri).165
G(ncremental deplo)-2.75 E(yment and backw)-.11 E(ards compatibility)
-.11 E(,)-.715 E(the RFC on ECN [RFC3168] speci\214es that routers may \
mark ECN-capable pack)93.6 215 Q(ets that)-.11 E -.11(wo)93.6 228 S
(uld otherwise ha).11 E .33 -.165(ve b)-.22 H
(een dropped, using the Congestion Experienced codepoint in the).165 E
(ECN \214eld.)93.6 241 Q(The use of ECN allo)5.5 E
(ws TCP to react to congestion while a)-.275 E -.22(vo)-.22 G
(iding unnecessary).22 E
(retransmissions and, in some cases, unnecessary retransmit timeouts.)
93.6 254 Q(Thus, using ECN has)5.5 E(se)93.6 267 Q -.165(ve)-.275 G
(ral bene\214ts:).165 E(1\) F)93.6 293 Q
(or short transfers, a TCP connection')-.165 E 2.75(sc)-.605 G
(ongestion windo)-2.75 E 2.75(wm)-.275 G(ay be small.)-2.75 E -.165(Fo)
5.5 G 2.75(re).165 G(xample, if)-2.915 E(the current windo)93.6 306 Q
2.75(wc)-.275 G(ontains only one pack)-2.75 E(et, and that pack)-.11 E
(et is dropped, TCP will ha)-.11 E .33 -.165(ve t)-.22 H(o).165 E -.11
(wa)93.6 319 S(it for a retransmit timeout to reco).11 E -.165(ve)-.165
G .88 -.44(r, r).165 H(educing its o).44 E -.165(ve)-.165 G
(rall throughput.).165 E(Similarly)5.5 E 2.75(,i)-.715 G 2.75(ft)-2.75 G
(he)-2.75 E(current windo)93.6 332 Q 2.75(wc)-.275 G(ontains only a fe)
-2.75 E 2.75(wp)-.275 G(ack)-2.75 E(ets and one of those pack)-.11 E
(ets is dropped, there might)-.11 E(not be enough duplicate ackno)93.6
345 Q(wledgements for a f)-.275 E
(ast retransmission, and the sender might)-.11 E(ha)93.6 358 Q .33 -.165
(ve t)-.22 H 2.75(ow).165 G(ait for a delay of se)-2.86 E -.165(ve)-.275
G(ral round-trip times using Limited T).165 E(ransmit [RFC3042].)-.385 E
-.44(Wi)93.6 371 S(th the use of ECN, short \215o).44 E(ws are less lik)
-.275 E(ely to ha)-.11 E .33 -.165(ve p)-.22 H(ack).165 E
(ets dropped, sometimes)-.11 E -.22(avo)93.6 384 S
(iding unnecessary delays or costly retransit timeouts.).22 E
(2\) While longer \215o)93.6 410 Q(ws may not see substantially impro)
-.275 E -.165(ve)-.165 G 2.75(dt).165 G(hroughput with the use of ECN,)
-2.75 E(the)93.6 423 Q 2.75(ye)-.165 G(xperience lo)-2.915 E
(wer loss. This may bene\214t TCP applications that are latenc)-.275 E
(y- and loss-)-.165 E(sensiti)93.6 436 Q -.165(ve)-.275 G 2.75(,b).165 G
(ecause of the a)-2.75 E -.22(vo)-.22 G(idance of retransmissions.).22 E
(RFC 3168 only speci\214es marking the Congestion Experienced codepoint\
 on TCP')93.6 462 Q 2.75(sd)-.605 G(ata)-2.75 E(pack)93.6 475 Q
(ets, and not on SYN and SYN/A)-.11 E(CK pack)-.44 E 2.75(ets. RFC)-.11
F(3168 speci\214es the ne)2.75 E(gotiation of the)-.165 E
(use of ECN between the tw)93.6 488 Q 2.75(oT)-.11 G
(CP end-points in the TCP SYN and SYN-A)-2.75 E(CK e)-.44 E(xchange,)
-.165 E(using \215ags in the TCP header)93.6 501 Q 5.5(.E)-.605 G
(rring on the side of being conserv)-5.5 E(ati)-.275 E -.165(ve)-.275 G
2.75(,R).165 G(FC 3168 does not)-2.75 E
(specify the use of ECN for the SYN/A)93.6 514 Q(CK pack)-.44 E
(et itself.)-.11 E(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, b)
.165 H(ecause of the high cost).44 E(to the TCP transfer of ha)93.6 527
Q(ving a SYN/A)-.22 E(CK pack)-.44 E
(et dropped, with the resulting retransmit)-.11 E
(timeout, this document speci\214es the use of ECN for the SYN/A)93.6
540 Q(CK pack)-.44 E(et itself.)-.11 E(This can be)5.5 E
(of great bene\214t to the TCP connection, a)93.6 553 Q -.22(vo)-.22 G
(iding the se).22 E -.165(ve)-.275 G(re penalty of a retransmit timeout)
.165 E
(for a connection that has not yet started placing a load on the netw)
93.6 566 Q 2.75(ork. The)-.11 F(sender of the)2.75 E(SYN/A)93.6 579 Q
(CK pack)-.44 E(et must respond to a report of an ECN-mark)-.11 E
(ed SYN/A)-.11 E(CK pack)-.44 E(et by)-.11 E
(reducing its initial congestion windo)93.6 592 Q 2.75(wf)-.275 G
(rom tw)-2.75 E(o, three, or four se)-.11 E(gments to one se)-.165 E
(gment,)-.165 E
(reducing the subsequent load from that connection on the netw)93.6 605
Q(ork.)-.11 E(The use of ECN for SYN/A)93.6 631 Q(CK pack)-.44 E
(ets has the follo)-.11 E(wing potential bene\214ts:)-.275 E(1\) A)93.6
644 Q -.22(vo)-.814 G(idance of a retransmit timeout;).22 E(2\) Impro)
93.6 657 Q -.165(ve)-.165 G
(ment in the throughput of short connections.).165 E
(This draft speci\214es ECN+, a modi\214cation to RFC 3168 to allo)93.6
683 Q 2.75(wT)-.275 G(CP SYN/A)-2.75 E(CK pack)-.44 E(ets to)-.11 E
(be ECN-Capable.)93.6 696 Q
(Section 2 contains the speci\214cation of the change, while Section 3)
5.5 E(Flo)72 733 Q 366.891(yd [P)-.11 F(age 4])-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E
(discusses some of the issues, and Section 4 discusses related w)93.6 85
Q 2.75(ork. Section)-.11 F 2.75(5c)2.75 G(ontains an)-2.75 E -.275(eva)
93.6 98 S(luation of the proposed change.).275 E/F1 11/Times-Bold@0 SF
(3.)72 124 Q/F2 14/Times-Bold@0 SF(Pr)5.5 E(oposal)-.252 E F0
(This section speci\214es the modi\214cation to RFC 3168 to allo)93.6
150 Q 2.75(wT)-.275 G(CP SYN/A)-2.75 E(CK pack)-.44 E(ets to be)-.11 E
2.75(ECN-Capable. W)93.6 163 R 2.75(eu)-.88 G(se the follo)-2.75 E
(wing terminology from RFC 3168:)-.275 E
(The ECN \214eld in the IP header:)93.6 189 Q 5.5(oC)93.6 202 S
(E: the Congestion Experienced codepoint; and)-5.5 E 5.5(oE)93.6 215 S
(CT)-5.5 E 2.75(:e)-.55 G(ither one of the tw)-2.75 E 2.75(oE)-.11 G
(CN-Capable T)-2.75 E(ransport codepoints.)-.385 E
(The ECN \215ags in the TCP header:)93.6 241 Q 5.5(oC)93.6 254 S
(WR: the Congestion W)-5.5 E(indo)-.44 E 2.75(wR)-.275 G
(educed \215ag; and)-2.75 E 5.5(oE)93.6 267 S(CE: the ECN-Echo \215ag.)
-5.5 E(ECN-setup pack)93.6 293 Q(ets:)-.11 E 5.5(oE)93.6 306 S
(CN-setup SYN pack)-5.5 E(et: a SYN pack)-.11 E
(et with the ECE and CWR \215ags;)-.11 E 5.5(oE)93.6 319 S
(CN-setup SYN-A)-5.5 E(CK pack)-.44 E(et: a SYN-A)-.11 E(CK pack)-.44 E
(et with ECE b)-.11 E(ut not CWR.)-.22 E
(RFC 3168 in Section 6.1.1. states that "A host MUST NO)93.6 345 Q 2.75
(Ts)-.44 G(et ECT on SYN or SYN-A)-2.75 E(CK)-.44 E(pack)93.6 358 Q
(ets." In this section, we specify that a TCP node MA)-.11 E 2.75(Yr)
-1.155 G(espond to an ECN-setup SYN)-2.75 E(pack)93.6 371 Q
(et by setting ECT in the responding ECN-setup SYN/A)-.11 E(CK pack)-.44
E(et, indicating to routers)-.11 E(that the SYN/A)93.6 384 Q(CK pack)
-.44 E(et is ECN-Capable.)-.11 E(This allo)5.5 E
(ws a congested router along the path to)-.275 E(mark the pack)93.6 397
Q(et instead of dropping the pack)-.11 E
(et as an indication of congestion.)-.11 E
(Assume that TCP node A transmits to TCP node B an ECN-setup SYN pack)
93.6 423 Q(et, indicating)-.11 E
(willingness to use ECN for this connection.)93.6 436 Q
(As speci\214ed by RFC 3168, if TCP node B is)5.5 E
(willing to use ECN, node B responds with an ECN-setup SYN-A)93.6 449 Q
(CK pack)-.44 E(et.)-.11 E -.88(Ta)93.6 475 S(ble 1 sho).88 E
(ws an interchange with the SYN/A)-.275 E(CK pack)-.44 E
(et dropped by a congested router)-.11 E(.)-.605 E(Node B w)93.6 488 Q
(aits for a retransmit timeout, and then retransmits the SYN/A)-.11 E
(CK pack)-.44 E(et.)-.11 E(Flo)72 733 Q 366.891(yd [P)-.11 F(age 5])
-.165 E 0 Cg EP
%%Page: 6 6
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E
(---------------------------------------------------------------)129.6
85 Q(TCP Node A)137.85 98 Q 46.75(Router TCP)35.75 F(Node B)2.75 E 33
(---------- ------)137.85 111 R(----------)49.5 E(ECN-setup SYN pack)
137.85 137 Q(et --->)-.11 E(ECN-setup SYN pack)228.6 150 Q(et --->)-.11
E(<--- ECN-setup SYN/A)198.35 176 Q(CK, possibly ECT)-.44 E
(3-second timer set)247.85 189 Q(SYN/A)192.85 202 Q(CK dropped)-.44 E(.)
41.25 E(.)275.35 215 Q(.)275.35 228 Q(3-second timer e)236.85 241 Q
(xpires)-.165 E(<--- ECN-setup SYN/A)212.1 254 Q(CK, not ECT)-.44 E
(<--- ECN-setup SYN/A)137.85 267 Q(CK)-.44 E(Data/A)137.85 280 Q
(CK --->)-.44 E(Data/A)261.6 293 Q(CK --->)-.44 E
(<--- Data \(one to four se)209.35 306 Q(gments\))-.165 E
(---------------------------------------------------------------)129.6
319 Q -.88(Ta)137.85 345 S(ble 1: SYN e).88 E(xchange with the SYN/A)
-.165 E(CK pack)-.44 E(et dropped.)-.11 E(If the SYN/A)93.6 377 Q
(CK pack)-.44 E(et is dropped in the netw)-.11 E
(ork, the TCP host \(node B\) responds by)-.11 E -.11(wa)93.6 390 S
(iting three seconds for the retransmit timer to e).11 E
(xpire [RFC2988].)-.165 E(If a SYN/A)5.5 E(CK pack)-.44 E(et)-.11 E(wit\
h the ECT codepoint is dropped, the TCP node SHOULD resend the SYN/A)
93.6 403 Q(CK pack)-.44 E(et)-.11 E(without the ECN-Capable codepoint.)
93.6 416 Q(\(Although we are not a)5.5 E -.11(wa)-.165 G(re of an).11 E
2.75(ym)-.165 G(iddlebox)-2.75 E(es that)-.165 E(drop SYN/A)93.6 429 Q
(CK pack)-.44 E
(ets that contain an ECN-Capable codepoint in the IP header)-.11 E 2.75
(,w)-.44 G 2.75(eh)-2.75 G -2.475 -.22(av e)-2.75 H
(learned to design our protocols defensi)93.6 442 Q -.165(ve)-.275 G
(ly in this re).165 E -.055(ga)-.165 G(rd [RFC3360].\)).055 E 1.76 -.88
(We n)93.6 468 T(ote that if syn-cookies were used by Node B in the e)
.88 E(xchange in T)-.165 E(able 1, TCP Node B)-.88 E -.11(wo)93.6 481 S
(uldn').11 E 2.75(ts)-.198 G(et a timer upon transmission of the SYN/A)
-2.75 E(CK pack)-.44 E(et [SYN-COOK].)-.11 E(In this case,)5.5 E
(if the SYN/A)93.6 494 Q(CK pack)-.44 E(et w)-.11 E
(as lost, the initiator \(Node A\) w)-.11 E(ould ha)-.11 E .33 -.165
(ve t)-.22 H 2.75(ot).165 G(imeout and)-2.75 E(retransmit the SYN pack)
93.6 507 Q(et in order to trigger another SYN-A)-.11 E(CK.)-.44 E -.88
(Ta)93.6 533 S(ble 2 sho).88 E(ws an interchange with the SYN/A)-.275 E
(CK pack)-.44 E(et sent as ECN-Capable, and ECN-)-.11 E(mark)93.6 546 Q
(ed instead of dropped at the congested router)-.11 E(.)-.605 E(Flo)72
733 Q 366.891(yd [P)-.11 F(age 6])-.165 E 0 Cg EP
%%Page: 7 7
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E
(---------------------------------------------------------------)129.6
85 Q(TCP Node A)137.85 98 Q 46.75(Router TCP)35.75 F(Node B)2.75 E 33
(---------- ------)137.85 111 R(----------)49.5 E(ECN-setup SYN pack)
137.85 137 Q(et --->)-.11 E(ECN-setup SYN pack)225.85 150 Q(et --->)-.11
E(<--- ECN-setup SYN/A)220.35 176 Q(CK, ECT)-.44 E
(<--- Sets CE on SYN/A)190.1 189 Q(CK)-.44 E(<--- ECN-setup SYN/A)137.85
202 Q(CK, CE)-.44 E(Data/A)137.85 228 Q(CK, ECN-Echo --->)-.44 E(Data/A)
231.35 241 Q(CK, ECN-Echo --->)-.44 E -.44(Wi)212.1 254 S(ndo).44 E 2.75
(wr)-.275 G(educed to one se)-2.75 E(gment.)-.165 E
(<--- Data, CWR \(one se)203.85 267 Q(gment only\))-.165 E
(---------------------------------------------------------------)129.6
280 Q -.88(Ta)137.85 306 S(ble 2: SYN e).88 E(xchange with the SYN/A)
-.165 E(CK pack)-.44 E(et mark)-.11 E(ed.)-.11 E(If the recei)93.6 338 Q
(ving node \(node A\) recei)-.275 E -.165(ve)-.275 G 2.75(saS).165 G
(YN/A)-2.75 E(CK pack)-.44 E(et that has been mark)-.11 E(ed by the)-.11
E(congested router)93.6 351 Q 2.75(,w)-.44 G
(ith the CE codepoint set, the recei)-2.75 E
(ving node MUST respond by setting)-.275 E
(the ECN-Echo \215ag in the TCP header of the responding A)93.6 364 Q
(CK pack)-.44 E 2.75(et. As)-.11 F(speci\214ed in)2.75 E
(RFC 3168, the recei)93.6 377 Q
(ving node continues to set the ECN-Echo \215ag in pack)-.275 E
(ets until it recei)-.11 E -.165(ve)-.275 G(s).165 E 2.75(ap)93.6 390 S
(ack)-2.75 E(et with the CWR \215ag set.)-.11 E
(When the sending node \(node B\) recei)93.6 416 Q -.165(ve)-.275 G 2.75
(st).165 G(he ECN-Echo pack)-2.75 E(et reporting the Congestion)-.11 E
(Experienced indication in the SYN/A)93.6 429 Q(CK pack)-.44 E
(et, the node MUST set the initial congestion)-.11 E(windo)93.6 442 Q
2.75(wt)-.275 G 2.75(oo)-2.75 G(ne se)-2.75 E(gment, instead of tw)-.165
E 2.75(os)-.11 G -.165(eg)-2.75 G(ments as allo).165 E
(wed by [RFC2581], or three or four)-.275 E(se)93.6 455 Q(gments allo)
-.165 E(wed by [RFC3390].)-.275 E(If the sending node \(node B\) w)5.5 E
(as going to use an initial)-.11 E(windo)93.6 468 Q 2.75(wo)-.275 G 2.75
(fo)-2.75 G(ne se)-2.75 E(gment, and recei)-.165 E -.165(ve)-.275 G 2.75
(sa).165 G 2.75(nE)-2.75 G(CN-Echo pack)-2.75 E
(et informing it of a Congestion)-.11 E
(Experienced indication on its SYN/A)93.6 481 Q(CK pack)-.44 E
(et, the sending node MA)-.11 E 2.75(Yc)-1.155 G(ontinue to send)-2.75 E
(with an initial windo)93.6 494 Q 2.75(wo)-.275 G 2.75(fo)-2.75 G(ne se)
-2.75 E(gment, without w)-.165 E(aiting for a retransmit timeout.)-.11 E
1.76 -.88(We n)5.5 H(ote).88 E(that this updates RFC 3168, which speci\
\214es that "the sending TCP MUST reset the)93.6 507 Q
(retransmit timer on recei)93.6 520 Q(ving the ECN-Echo pack)-.275 E
(et when the congestion windo)-.11 E 2.75(wi)-.275 G 2.75(so)-2.75 G
(ne.")-2.75 E(As speci\214ed by RFC 3168, the sending node \(node B\) a\
lso sets the CWR \215ag in the TCP)93.6 533 Q(header of the ne)93.6 546
Q(xt data pack)-.165 E(et sent, to ackno)-.11 E
(wledge its receipt of and reaction to the ECN-)-.275 E(Echo \215ag.)
93.6 559 Q(If the data transfer in T)93.6 585 Q
(able 2 is entirely from Node A to Node B, then data pack)-.88 E
(ets from)-.11 E
(Node A continue to set the ECN-Echo \215ag in data pack)93.6 598 Q
(ets, w)-.11 E(aiting for the CWR \215ag from)-.11 E(Node B ackno)93.6
611 Q(wledging a response to the ECN-Echo \215ag.)-.275 E/F1 11
/Times-Bold@0 SF(4.)72 650 Q/F2 14/Times-Bold@0 SF(Discussion)5.5 E F0
(Moti)93.6 676 Q -.275(va)-.275 G(tion:).275 E
(The rationale for the proposed change is the follo)93.6 689 Q 2.75
(wing. When)-.275 F(node B recei)2.75 E -.165(ve)-.275 G 2.75(saT).165 G
(CP SYN)-2.75 E(Flo)72 733 Q 366.891(yd [P)-.11 F(age 7])-.165 E 0 Cg EP
%%Page: 8 8
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(pack)93.6 85 Q(et with ECN-Echo bit set in the TCP header)
-.11 E 2.75(,t)-.44 G(his indicates that node A is ECN-capable.)-2.75 E
(If node B is also ECN-capable, there are no obstacles to immediately s\
etting one of the)93.6 98 Q
(ECN-Capable codepoints in the IP header in the responding TCP SYN/A)
93.6 111 Q(CK pack)-.44 E(et.)-.11 E(There can be a great bene\214t in \
setting an ECN-capable codepoint in SYN/A)93.6 137 Q(CK pack)-.44 E
(ets, as is)-.11 E(discussed further in Section 4.)93.6 150 Q
(Congestion is most lik)5.5 E(ely to occur in the serv)-.11 E(er)-.165 E
(-to-client)-.22 E 2.75(direction. As)93.6 163 R 2.75(ar)2.75 G
(esult, setting an ECN-capable codepoint in SYN/A)-2.75 E(CK pack)-.44 E
(ets can reduce)-.11 E(the occurrence of three-second retransmit timeou\
ts resulting from the drop of SYN/A)93.6 176 Q(CK)-.44 E(pack)93.6 189 Q
(ets.)-.11 E(Flooding attacks:)93.6 215 Q
(Setting an ECN-Capable codepoint in the responding TCP SYN/A)93.6 228 Q
(CK pack)-.44 E(ets does not raise)-.11 E(an)93.6 241 Q 2.75(yn)-.165 G
-.165(ove)-2.75 G 2.75(ls).165 G(ecurity vulnerabilities.)-2.75 E -.165
(Fo)5.5 G 2.75(re).165 G(xample, pro)-2.915 E -.22(vo)-.165 G(king serv)
.22 E(ers or hosts to send)-.165 E(SYN/A)93.6 254 Q(CK pack)-.44 E
(ets to a third party in order to perform a "SYN/A)-.11 E
(CK \215ood" attack w)-.44 E(ould be)-.11 E(highly inef)93.6 267 Q 2.75
(\214cient. Third)-.275 F(parties w)2.75 E
(ould immediately drop such pack)-.11 E(ets, since the)-.11 E 2.75(yw)
-.165 G(ould)-2.86 E(kno)93.6 280 Q 2.75(wt)-.275 G(hat the)-2.75 E 2.75
(yd)-.165 G(idn')-2.75 E 2.75(tg)-.198 G(enerate the TCP SYN pack)-2.75
E(ets in the \214rst place.)-.11 E(Moreo)5.5 E -.165(ve)-.165 G .88 -.44
(r, s).165 H(uch).44 E(SYN/A)93.6 293 Q(CK attacks w)-.44 E(ould ha)-.11
E .33 -.165(ve t)-.22 H(he same signatures as the e).165 E
(xisting TCP SYN attacks.)-.165 E(Pro)93.6 306 Q -.22(vo)-.165 G
(king serv).22 E(ers or hosts to reply with SYN/A)-.165 E(CK pack)-.44 E
(ets in order to congest a certain link)-.11 E -.11(wo)93.6 319 S
(uld also be highly inef).11 E(\214cient because SYN/A)-.275 E(CK pack)
-.44 E(ets are small in size.)-.11 E(Ho)93.6 345 Q(we)-.275 E -.165(ve)
-.275 G .88 -.44(r, t).165 H(he addition of ECN-Capability to SYN/A).44
E(CK pack)-.44 E(ets could allo)-.11 E 2.75(wS)-.275 G(YN/A)-2.75 E(CK)
-.44 E(pack)93.6 358 Q(ets to persist for more hops along a netw)-.11 E
(ork path before being dropped, thus adding)-.11 E(some)93.6 371 Q
(what to the ability of a SYN/A)-.275 E(CK attack to \215ood a netw)-.44
E(ork link.)-.11 E(The TCP SYN pack)93.6 397 Q(et:)-.11 E(There are se)
93.6 410 Q -.165(ve)-.275 G(ral reasons wh).165 E 2.75(ya)-.055 G 2.75
(nE)-2.75 G(CN-Capable codepoint MUST NO)-2.75 E 2.75(Tb)-.44 G 2.75(es)
-2.75 G(et in the IP)-2.75 E(header of the initiating TCP SYN pack)93.6
423 Q 2.75(et. First,)-.11 F(when the TCP SYN pack)2.75 E
(et is sent, there are)-.11 E
(no guarantees that the other TCP endpoint \(node B in T)93.6 436 Q
(able 2\) is ECN-capable, or that it)-.88 E -.11(wo)93.6 449 S
(uld be able to understand and react if the ECN CE codepoint w).11 E
(as set by a congested)-.11 E(router)93.6 462 Q(.)-.605 E
(Second, the ECN-Capable codepoint in TCP SYN pack)93.6 488 Q
(ets could be misused by malicious)-.11 E(clients to `impro)93.6 501 Q
-.165(ve)-.165 G 2.75('t).165 G(he well-kno)-2.75 E
(wn TCP SYN attack. By setting an ECN-Capable codepoint)-.275 E
(in TCP SYN pack)93.6 514 Q
(ets, a malicious host might be able to inject a lar)-.11 E
(ge number of TCP SYN)-.198 E(pack)93.6 527 Q
(ets through a potentially congested ECN-enabled router)-.11 E 2.75(,c)
-.44 G(ongesting it e)-2.75 E -.165(ve)-.275 G 2.75(nf).165 G(urther)
-2.75 E(.)-.605 E -.165(Fo)93.6 553 S 2.75(rb).165 G
(oth these reasons, we continue the restriction that the TCP SYN pack)
-2.75 E(et MUST NO)-.11 E(T)-.44 E(ha)93.6 566 Q .33 -.165(ve t)-.22 H
(he ECN-Capable codepoint in the IP header set.).165 E(Backw)93.6 592 Q
(ards compatibility:)-.11 E(In order for TCP node B to send a SYN/A)93.6
605 Q(CK pack)-.44 E(et as ECN-Capable, node B must ha)-.11 E -.165(ve)
-.22 G(recei)93.6 618 Q -.165(ve)-.275 G 2.75(da).165 G 2.75(nE)-2.75 G
(CN-setup SYN pack)-2.75 E(et from node A.)-.11 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(sp)-2.75 G
(ossible that node A)-2.75 E(supports ECN, b)93.6 631 Q
(ut either ignores the CE codepoint on recei)-.22 E -.165(ve)-.275 G
2.75(dS).165 G(YN/A)-2.75 E(CK pack)-.44 E(ets, or)-.11 E(ignores SYN/A)
93.6 644 Q(CK pack)-.44 E(ets with the ECT or CE codepoint set.)-.11 E
(If the TCP sender ignores the)5.5 E(CE codepoint on recei)93.6 657 Q
-.165(ve)-.275 G 2.75(dS).165 G(YN/A)-2.75 E(CK pack)-.44 E(ets, this w)
-.11 E(ould mean that the TCP connection)-.11 E -.11(wo)93.6 670 S
(uld not respond to this congestion indication.).11 E
(As discussed in Section 2 under)5.5 E("Backw)93.6 683 Q
(ards compatibility", this w)-.11 E
(ould not be an insurmountable problem.)-.11 E(It w)5.5 E(ould mean)-.11
E(that the sender of the SYN/A)93.6 696 Q(CK pack)-.44 E(et w)-.11 E
(ould not reduce the initial congestion windo)-.11 E 2.75(wf)-.275 G
(rom)-2.75 E(Flo)72 733 Q 366.891(yd [P)-.11 F(age 8])-.165 E 0 Cg EP
%%Page: 9 9
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(tw)93.6 85 Q(o, three, or four se)-.11 E(gments do)-.165 E
(wn to one se)-.275 E(gment, as it should.)-.165 E(Ho)5.5 E(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, t).165 H(he TCP sender).44 E -.11(wo)93.6
98 S(uld still respond correctly to an).11 E 2.75(ys)-.165 G
(ubsequent CE indications on data pack)-2.75 E(ets later on in the)-.11
E(connection.)93.6 111 Q(It is also possible that in some older TCP imp\
lementation, the TCP sender ignores)93.6 137 Q(SYN/A)93.6 150 Q(CK pack)
-.44 E(ets with the ECT or CE codepoint set.)-.11 E(This w)5.5 E
(ould result in a delay in)-.11 E(connection set-up for that TCP connec\
tion, with the TCP sender re-sending the SYN pack)93.6 163 Q(et)-.11 E
(after a retransmit timeout.)93.6 176 Q(SYN/A)93.6 202 Q(CK pack)-.44 E
(ets and pack)-.11 E(et size:)-.11 E(There are a number of router b)93.6
215 Q(uf)-.22 E(fer architectures that ha)-.275 E .33 -.165(ve s)-.22 H
(maller dropping rates for small).165 E(\(SYN\) pack)93.6 228 Q
(ets than for lar)-.11 E(ge \(data\) pack)-.198 E 2.75(ets. F)-.11 F
(or e)-.165 E(xample, for a Drop T)-.165 E(ail queue in units of)-.88 E
(pack)93.6 241 Q(ets, where each pack)-.11 E(et tak)-.11 E
(es a single slot in the b)-.11 E(uf)-.22 E(fer re)-.275 E -.055(ga)
-.165 G(rdless of pack).055 E(et size, small)-.11 E(and lar)93.6 254 Q
(ge pack)-.198 E(ets are equally lik)-.11 E(ely to be dropped.)-.11 E
(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H(or a Drop T)
.44 E(ail queue in units)-.88 E(of bytes, small pack)93.6 267 Q
(ets are less lik)-.11 E(ely to be dropped than are lar)-.11 E(ge ones.)
-.198 E(Similarly)5.5 E 2.75(,f)-.715 G(or RED)-2.75 E(in pack)93.6 280
Q(et mode, small and lar)-.11 E(ge pack)-.198 E(ets are equally lik)-.11
E(ely to be dropped or mark)-.11 E(ed, while)-.11 E
(for RED in byte mode, a pack)93.6 293 Q(et')-.11 E 2.75(sc)-.605 G
(hance of being dropped or mark)-2.75 E(ed is proportional to the)-.11 E
(pack)93.6 306 Q(et size in bytes.)-.11 E -.165(Fo)93.6 332 S 2.75(rac)
.165 G(ongested router with an A)-2.75 E
(QM mechanism in byte mode, where a pack)-.605 E(et')-.11 E 2.75(sc)
-.605 G(hance of)-2.75 E(being dropped or mark)93.6 345 Q
(ed is proportional to the pack)-.11 E
(et size in bytes, the drop or marking rate)-.11 E(for TCP SYN/A)93.6
358 Q(CK pack)-.44 E(ets should generally be lo)-.11 E 4.18 -.715(w. I)
-.275 H 2.75(nt).715 G(his case, the bene\214t of making)-2.75 E(SYN/A)
93.6 371 Q(CK pack)-.44 E(ets ECN-Capable should be similarly moderate.)
-.11 E(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H
(or a congested).44 E(router with a Drop T)93.6 384 Q
(ail queue in units of pack)-.88 E(ets or with an A)-.11 E
(QM mechanism in pack)-.605 E(et)-.11 E
(mode, and with no priority queueing for smaller pack)93.6 397 Q
(ets, small and lar)-.11 E(ge pack)-.198 E(ets should)-.11 E(ha)93.6 410
Q .33 -.165(ve t)-.22 H(he same probability of being dropped or mark)
.165 E 2.75(ed. In)-.11 F(such a case, making SYN/A)2.75 E(CK)-.44 E
(pack)93.6 423 Q(ets ECN-Capable should be of signi\214cant bene\214t.)
-.11 E 1.76 -.88(We b)93.6 449 T(elie).88 E .33 -.165(ve t)-.275 H
(hat there are a wide range of beha).165 E(viors in the real w)-.22 E
(orld in terms of the drop or)-.11 E(mark beha)93.6 462 Q
(vior at routers as a function of pack)-.22 E(et size [T)-.11 E
(ools] \(Section 10\).)-.88 E 1.76 -.88(We n)5.5 H(ote that all of).88 E
(these alternati)93.6 475 Q -.165(ve)-.275 G 2.75(sl).165 G(isted abo)
-2.75 E .33 -.165(ve a)-.165 H(re a).165 E -.275(va)-.22 G
(ilable in the NS simulator \(Drop T).275 E(ail queues are by)-.88 E
(def)93.6 488 Q(ault in units of pack)-.11 E(ets, while the def)-.11 E
(ault for RED queue management has been changed)-.11 E(from pack)93.6
501 Q(et mode to byte mode\).)-.11 E(Response to ECN-marking of SYN/A)
93.6 527 Q(CK pack)-.44 E(ets:)-.11 E(One question is wh)93.6 540 Q 2.75
(yT)-.055 G(CP SYN/A)-2.75 E(CK pack)-.44 E(ets should be treated dif)
-.11 E(ferently from other)-.275 E(pack)93.6 553 Q
(ets in terms of the pack)-.11 E(et sender')-.11 E 2.75(sr)-.605 G
(esponse to an ECN-mark)-2.75 E(ed pack)-.11 E 2.75(et. Section)-.11 F
2.75(5o)2.75 G(f)-2.75 E(RFC 3168 speci\214es the follo)93.6 566 Q
(wing:)-.275 E
("Upon the receipt by an ECN-Capable transport of a single CE pack)93.6
592 Q(et, the congestion)-.11 E(control algorithms follo)93.6 605 Q
(wed at the end-systems MUST be essentially the same as the)-.275 E
(congestion control response to a *single* dropped pack)93.6 618 Q 2.75
(et. F)-.11 F(or e)-.165 E(xample, for ECN-Capable)-.165 E
(TCP the source TCP is required to halv)93.6 631 Q 2.75(ei)-.165 G
(ts congestion windo)-2.75 E 2.75(wf)-.275 G(or an)-2.75 E 2.75(yw)-.165
G(indo)-2.75 E 2.75(wo)-.275 G 2.75(fd)-2.75 G(ata)-2.75 E
(containing either a pack)93.6 644 Q(et drop or an ECN indication.")-.11
E(In particular)93.6 670 Q 2.75(,S)-.44 G(ection 6.1.2 of RFC 3168 spec\
i\214es that when the TCP congestion windo)-2.75 E(w)-.275 E
(consists of a single pack)93.6 683 Q(et and that pack)-.11 E
(et is ECN-mark)-.11 E(ed in the netw)-.11 E(ork, then the sender)-.11 E
(must reduce the sending rate belo)93.6 696 Q 2.75(wo)-.275 G(ne pack)
-2.75 E(et per round-trip time, by w)-.11 E(aiting for one R)-.11 E
-.198(TO)-.66 G(Flo)72 733 Q 366.891(yd [P)-.11 F(age 9])-.165 E 0 Cg EP
%%Page: 10 10
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(before sending another pack)93.6 85 Q 2.75(et. If)-.11 F
(the R)2.75 E .396 -.198(TO w)-.66 H(as set to the a).088 E -.165(ve)
-.22 G(rage round-trip time, this w).165 E(ould)-.11 E
(result in halving the sending rate; because the R)93.6 98 Q .396 -.198
(TO i)-.66 H 2.75(si).198 G 2.75(nf)-2.75 G(act lar)-2.86 E
(ger than the a)-.198 E -.165(ve)-.22 G(rage round-).165 E
(trip time, the sending rate is reduced to less than half of its pre)
93.6 111 Q(vious v)-.275 E(alue.)-.275 E(TCP')93.6 137 Q 2.75(sc)-.605 G
(ongestion control response to the *dropping* of a SYN/A)-2.75 E
(CK pack)-.44 E(et is to w)-.11 E(ait a)-.11 E(def)93.6 150 Q
(ault time before sending another pack)-.11 E 2.75(et. This)-.11 F
(document ar)2.75 E(gues that ECN gi)-.198 E -.165(ve)-.275 G 2.75(se)
.165 G(nd-)-2.75 E(systems a wider range of possible responses to the *\
marking* of a SYN/A)93.6 163 Q(CK pack)-.44 E(et, and)-.11 E(that w)93.6
176 Q(aiting a def)-.11 E(ault time before sending a data pack)-.11 E
(et is not the desired response.)-.11 E(On the conserv)93.6 202 Q(ati)
-.275 E .33 -.165(ve e)-.275 H(nd, one could assume an ef).165 E(fecti)
-.275 E .33 -.165(ve c)-.275 H(ongestion windo).165 E 2.75(wo)-.275 G
2.75(fo)-2.75 G(ne pack)-2.75 E(et for)-.11 E(the SYN/A)93.6 215 Q
(CK pack)-.44 E(et, and respond to an ECN-mark)-.11 E(ed SYN/A)-.11 E
(CK pack)-.44 E(et by reducing the)-.11 E(sending rate to one pack)93.6
228 Q(et e)-.11 E -.165(ve)-.275 G(ry tw).165 E 2.75(or)-.11 G
(ound-trip times.)-2.75 E(As an approximation, the TCP end-)5.5 E(node \
could measure the round-trip time T between the sending of the SYN/A)
93.6 241 Q(CK pack)-.44 E(et and)-.11 E(the receipt of the ackno)93.6
254 Q(wledgement, and reply to the ackno)-.275 E
(wledgement of the ECN-mark)-.275 E(ed)-.11 E(SYN/A)93.6 267 Q(CK pack)
-.44 E(et by w)-.11 E(aiting T seconds before sending a data pack)-.11 E
(et.)-.11 E(Ho)93.6 293 Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, w).165
H 2.75(en).44 G(ote that for an ECN-mark)-2.75 E(ed SYN/A)-.11 E
(CK pack)-.44 E(et, halving the *congestion)-.11 E(windo)93.6 306 Q(w* \
is not the same as halving the *sending rate*; there is no `sending rat\
e' associated)-.275 E(with an ECN-Capable SYN/A)93.6 319 Q(CK pack)-.44
E(et, as such pack)-.11 E(ets are only sent as the \214rst pack)-.11 E
(et in a)-.11 E(connection from that host.)93.6 332 Q(Further)5.5 E 2.75
(,ar)-.44 G(outer')-2.75 E 2.75(sm)-.605 G(arking of a SYN/A)-2.75 E
(CK pack)-.44 E(et is not af)-.11 E(fected)-.275 E(by an)93.6 345 Q 2.75
(yp)-.165 G(ast history of that connection.)-2.75 E
(Adding ECN-Capability to SYN/A)93.6 371 Q(CK pack)-.44 E(ets allo)-.11
E(ws the simple response of setting the)-.275 E
(initial congestion windo)93.6 384 Q 2.75(wt)-.275 G 2.75(oo)-2.75 G
(ne pack)-2.75 E(et, instead of its allo)-.11 E(wed def)-.275 E(ault v)
-.11 E(alue of tw)-.275 E(o, three, or)-.11 E(four pack)93.6 397 Q
(ets, with the host proceeding with a cautious sending rate of one pack)
-.11 E(et per round-)-.11 E(trip time.)93.6 410 Q(If that pack)5.5 E
(et is ECN-mark)-.11 E(ed or dropped, then the sender will w)-.11 E
(ait an R)-.11 E .396 -.198(TO b)-.66 H(efore).198 E
(sending another pack)93.6 423 Q 2.75(et. This)-.11 F(document ar)2.75 E
(gues that this approach is useful to users, with no)-.198 E
(dangers of congestion collapse or of starv)93.6 436 Q
(ation of competing traf)-.275 E 2.75(\214c. This)-.275 F
(is discussed in)2.75 E(more detail belo)93.6 449 Q 2.75(wi)-.275 G 2.75
(nS)-2.75 G(ection 5.2.)-2.75 E 1.76 -.88(We n)93.6 475 T(ote that if t\
he data transfer is entirely from Node A to Node B, then there is no ef)
.88 E(fecti)-.275 E -.165(ve)-.275 G(dif)93.6 488 Q
(ference between the tw)-.275 E 2.75(op)-.11 G
(ossible responses to an ECN-mark)-2.75 E(ed SYN/A)-.11 E(CK pack)-.44 E
(et outlined)-.11 E(abo)93.6 501 Q -.165(ve)-.165 G 5.5(.I).165 G 2.75
(ne)-5.5 G(ither case, Node B sends no data pack)-2.75 E
(ets, only sending ackno)-.11 E(wledgement pack)-.275 E(ets)-.11 E
(in response to recei)93.6 514 Q -.165(ve)-.275 G 2.75(dd).165 G
(ata pack)-2.75 E(ets.)-.11 E/F1 11/Times-Bold@0 SF(5.)72 540 Q/F2 14
/Times-Bold@0 SF(Related W)5.5 E(ork)-1.05 E F0
(The addition of ECN-capability to TCP')93.6 566 Q 2.75(sS)-.605 G(YN/A)
-2.75 E(CK pack)-.44 E(ets w)-.11 E(as proposed in [ECN+].)-.11 E(The)
5.5 E(paper includes an e)93.6 579 Q(xtensi)-.165 E .33 -.165(ve s)-.275
H(et of simulation and testbed e).165 E(xperiments to e)-.165 E -.275
(va)-.275 G(luate the ef).275 E(fects)-.275 E(of the proposal, using se)
93.6 592 Q -.165(ve)-.275 G(ral Acti).165 E .33 -.165(ve Q)-.275 H
(ueue Management \(A).165 E(QM\) mechanisms, including)-.605 E(Random E\
arly Detection \(RED\) [RED], Random Exponential Marking \(REM\) [REM],\
 and)93.6 605 Q(Proportional Inte)93.6 618 Q(grator \(PI\) [PI].)-.165 E
(The performance measures were the end-to-end response)5.5 E
(times for each request/response pair)93.6 631 Q 2.75(,a)-.44 G
(nd the aggre)-2.75 E -.055(ga)-.165 G
(te throughput on the bottleneck link.).055 E
(The end-to-end response time w)93.6 644 Q
(as computed as the time from the moment when the request)-.11 E
(for the \214le is sent to the serv)93.6 657 Q(er)-.165 E 2.75(,u)-.44 G
(ntil that \214le is successfully do)-2.75 E(wnloaded by the client.)
-.275 E(The measurements from [ECN+] sho)93.6 683 Q 2.75(wt)-.275 G
(hat setting an ECN-Capable codepoint in the IP)-2.75 E(pack)93.6 696 Q
(et header in TCP SYN/A)-.11 E(CK pack)-.44 E(ets systematically impro)
-.11 E -.165(ve)-.165 G 2.75(sp).165 G(erformance with all)-2.75 E(Flo)
72 733 Q 361.391(yd [P)-.11 F(age 10])-.165 E 0 Cg EP
%%Page: 11 11
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E -.275(eva)93.6 85 S(luated A).275 E(QM schemes.)-.605 E
(When SYN/A)5.5 E(CK pack)-.44 E(ets at a congested router are ECN-mark)
-.11 E(ed)-.11 E(instead of dropped, this can a)93.6 98 Q -.22(vo)-.22 G
(id a long initial retransmit timeout, impro).22 E(ving the response)
-.165 E(time for the af)93.6 111 Q(fected \215o)-.275 E 2.75(wd)-.275 G
(ramatically)-2.75 E(.)-.715 E([ECN+] sho)93.6 137 Q
(ws that the impact on aggre)-.275 E -.055(ga)-.165 G
(te throughput can also be quite signi\214cant, because).055 E
(marking SYN A)93.6 150 Q(CK pack)-.44 E(ets can pre)-.11 E -.165(ve)
-.275 G(nt lar).165 E(ger \215o)-.198 E(ws from suf)-.275 E
(fering long timeouts before)-.275 E(being "admitted" into the netw)93.6
163 Q 2.75(ork. In)-.11 F
(addition, the testbed measurements from [ECN+])2.75 E(sho)93.6 176 Q
2.75(wt)-.275 G(hat web serv)-2.75 E
(ers setting the ECN-Capable codepoint in TCP SYN/A)-.165 E(CK pack)-.44
E(ets could)-.11 E(serv)93.6 189 Q 2.75(em)-.165 G(ore requests.)-2.75 E
(As a \214nal step, [ECN+] e)93.6 215 Q(xplores the co-e)-.165 E
(xistence of \215o)-.165 E(ws that do and don')-.275 E 2.75(ts)-.198 G
(et the ECN-)-2.75 E(capable codepoint in TCP SYN/A)93.6 228 Q(CK pack)
-.44 E 2.75(ets. The)-.11 F(results in [ECN+] sho)2.75 E 2.75(wt)-.275 G
(hat both types)-2.75 E(of \215o)93.6 241 Q(ws can coe)-.275 E
(xist, with some performance de)-.165 E(gradation for \215o)-.165 E
(ws that don')-.275 E 2.75(tu)-.198 G(se ECN+.)-2.75 E(Flo)93.6 254 Q
(ws that do use ECN+ impro)-.275 E .33 -.165(ve t)-.165 H
(heir end-to-end performance.).165 E(At the same time, the)5.5 E
(performance de)93.6 267 Q(gradation for \215o)-.165 E(ws that don')
-.275 E 2.75(tu)-.198 G(se ECN+, as a result of the \215o)-2.75 E
(ws that do use)-.275 E(ECN+, increases as a greater fraction of \215o)
93.6 280 Q(ws use ECN+.)-.275 E/F1 11/Times-Bold@0 SF(6.)72 306 Q/F2 14
/Times-Bold@0 SF -.28(Pe)5.5 G(rf).28 E(ormance Ev)-.35 E(aluation)-.14
E F1(6.1.)72 332 Q/F3 13/Times-Bold@0 SF
(The Costs and Bene\214t of Adding ECN-Capability)5.5 E F0([ECN+] e)93.6
358 Q
(xplores the costs and bene\214ts of adding ECN-Capability to SYN/A)
-.165 E(CK pack)-.44 E(ets)-.11 E(with both simulations and e)93.6 371 Q
2.75(xperiments. The)-.165 F(addition of ECN-capability to SYN/A)2.75 E
(CK)-.44 E(pack)93.6 384 Q(ets could be of signi\214cant bene\214t for \
those ECN connections that w)-.11 E(ould ha)-.11 E .33 -.165(ve h)-.22 H
(ad the).165 E(SYN/A)93.6 397 Q(CK pack)-.44 E(et dropped in the netw)
-.11 E(ork, and for which the ECN-Capability w)-.11 E(ould allo)-.11 E
(w)-.275 E(the SYN/A)93.6 410 Q(CK to be mark)-.44 E
(ed rather than dropped.)-.11 E(The percent of SYN/A)93.6 436 Q(CK pack)
-.44 E(ets on a link can be quite high. In particular)-.11 E 2.75(,m)
-.44 G(easurements)-2.75 E(on links dominated by web traf)93.6 449 Q
(\214c indicate that 15-20% of the pack)-.275 E(ets can be SYN/A)-.11 E
(CK)-.44 E(pack)93.6 462 Q(ets [SCJO01].)-.11 E
(The bene\214t of adding ECN-capability to SYN/A)93.6 488 Q(CK pack)-.44
E(ets depends in part on the size of)-.11 E(the data transfer)93.6 501 Q
5.5(.T)-.605 G(he drop of a SYN/A)-5.5 E(CK pack)-.44 E
(et can increase the do)-.11 E(wnload time of a short)-.275 E(\214le by\
 an order of magnitude, by requiring a three-second retransmit timeout.)
93.6 514 Q -.165(Fo)5.5 G 2.75(rl).165 G(onger)-2.75 E(-)-.22 E(li)93.6
527 Q -.165(ve)-.275 G 2.75<648d>.165 G -.275(ow)-2.75 G(s, the ef).275
E(fect of a dropped SYN/A)-.275 E(CK pack)-.44 E(et on \214le do)-.11 E
(wnload time is less dramatic.)-.275 E(Ho)93.6 540 Q(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, e).165 H -.165(ve).165 G 2.75(nf).165 G
(or longer)-2.75 E(-li)-.22 E -.165(ve)-.275 G 2.75<648d>.165 G -.275
(ow)-2.75 G(s, the addition of ECN-capability to SYN/A).275 E(CK pack)
-.44 E(ets)-.11 E(can impro)93.6 553 Q .33 -.165(ve t)-.165 H(he f).165
E(airness among long-li)-.11 E -.165(ve)-.275 G 2.75<648d>.165 G -.275
(ow)-2.75 G(s, as ne).275 E(wly-arri)-.275 E(ving \215o)-.275 E(ws w)
-.275 E(ould be less)-.11 E(lik)93.6 566 Q(ely to ha)-.11 E .33 -.165
(ve t)-.22 H 2.75(ow).165 G(ait for retransmit timeouts.)-2.86 E
(One question that arises is what fraction of connections w)93.6 592 Q
(ould see the bene\214t from making)-.11 E(SYN/A)93.6 605 Q(CK pack)-.44
E(ets ECN-capable, in a particular scenario?)-.11 E(Speci\214cally:)5.5
E(\(1\) What fraction of arri)93.6 631 Q(ving SYN/A)-.275 E(CK pack)-.44
E(ets are dropped at the congested router when)-.11 E(the SYN/A)93.6 644
Q(CK pack)-.44 E(ets are not ECN-capable?)-.11 E(\(2\) Of those SYN/A)
93.6 670 Q(CK pack)-.44 E
(ets that are dropped, what fraction of those drops w)-.11 E(ould ha)
-.11 E -.165(ve)-.22 G(been ECN-marks instead of drops if the SYN/A)93.6
683 Q(CK pack)-.44 E(ets had been ECN-capable?)-.11 E(Flo)72 733 Q
361.391(yd [P)-.11 F(age 11])-.165 E 0 Cg EP
%%Page: 12 12
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E 1.76 -.88(To a)93.6 85 T
(nswer \(1\), it is necessary to consider not only the le).88 E -.165
(ve)-.275 G 2.75(lo).165 G 2.75(fc)-2.75 G(ongestion b)-2.75 E
(ut also the queue)-.22 E(architecture at the congested link.)93.6 98 Q
(As described in Section 3 abo)5.5 E -.165(ve)-.165 G 2.75(,f).165 G
(or some queue)-2.75 E(architectures small pack)93.6 111 Q
(ets are less lik)-.11 E(ely to be dropped than lar)-.11 E(ge ones.)
-.198 E(In such an)5.5 E(en)93.6 124 Q(vironment, SYN/A)-.44 E(CK pack)
-.44 E(ets w)-.11 E(ould ha)-.11 E .33 -.165(ve l)-.22 H -.275(ow).165 G
(er pack).275 E(et drop rates; question \(1\) could not)-.11 E
(necessarily be inferred from the o)93.6 137 Q -.165(ve)-.165 G
(rall pack).165 E(et drop rate, b)-.11 E(ut could be answered by)-.22 E
(measuring the drop rate for SYN/A)93.6 150 Q(CK pack)-.44 E
(ets directly)-.11 E 5.5(.I)-.715 G 2.75(ns)-5.5 G(uch an en)-2.75 E
(vironment, adding)-.44 E(ECN-capability to SYN/A)93.6 163 Q(CK pack)
-.44 E(ets w)-.11 E(ould be of less dramatic bene\214t than in en)-.11 E
(vironments)-.44 E(where all pack)93.6 176 Q(ets are equally lik)-.11 E
(ely to be dropped re)-.11 E -.055(ga)-.165 G(rdless of pack).055 E
(et size.)-.11 E(As question \(2\) implies, e)93.6 202 Q -.165(ve)-.275
G 2.75(ni).165 G 2.75(fa)-2.75 G(ll of the SYN/A)-2.75 E(CK pack)-.44 E
(ets were ECN-capable, there could)-.11 E(still be some SYN/A)93.6 215 Q
(CK pack)-.44 E(ets dropped instead of mark)-.11 E
(ed at the congested link; the full)-.11 E(answer to question \(2\) dep\
ends on the details of the queue management mechanism at the)93.6 228 Q
(router)93.6 241 Q 5.5(.I)-.605 G 2.75(fc)-5.5 G(ongestion is suf)-2.75
E(\214ciently bad, and the queue management mechanism cannot)-.275 E
(pre)93.6 254 Q -.165(ve)-.275 G(nt the b).165 E(uf)-.22 E(fer from o)
-.275 E -.165(ve)-.165 G(r\215o).165 E(wing, then SYN/A)-.275 E(CK pack)
-.44 E(ets will be dropped rather than)-.11 E(mark)93.6 267 Q(ed upon b)
-.11 E(uf)-.22 E(fer o)-.275 E -.165(ve)-.165 G(r\215o).165 E 2.75(ww)
-.275 G(hether or not the)-2.75 E 2.75(ya)-.165 G(re ECN-capable.)-2.75
E -.165(Fo)93.6 293 S 2.75(rs).165 G(ome A)-2.75 E
(QM mechanisms, ECN-capable pack)-.605 E(ets are mark)-.11 E
(ed instead of dropped an)-.11 E 2.75(yt)-.165 G(ime)-2.75 E
(this is possible, that is, an)93.6 306 Q 2.75(yt)-.165 G(ime the b)
-2.75 E(uf)-.22 E(fer is not yet full.)-.275 E -.165(Fo)5.5 G 2.75(ro)
.165 G(ther A)-2.75 E(QM mechanisms)-.605 E(ho)93.6 319 Q(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, s).165 H
(uch as the RED mechanism as recommended in [RED], pack).44 E
(ets are dropped rather)-.11 E(than mark)93.6 332 Q(ed when the pack)
-.11 E(et drop/mark rate e)-.11 E
(xceeds a certain threshold, e.g., 10%, e)-.165 E -.165(ve)-.275 G 2.75
(ni).165 G(f)-2.75 E(the pack)93.6 345 Q(ets are ECN-capable.)-.11 E
-.165(Fo)5.5 G 2.75(rar).165 G(outer with such an A)-2.75 E
(QM mechanism, when congestion)-.605 E(is suf)93.6 358 Q(\214ciently se)
-.275 E -.165(ve)-.275 G(re to cause a high drop/mark rate, some SYN/A)
.165 E(CK pack)-.44 E(ets w)-.11 E(ould be)-.11 E
(dropped instead of mark)93.6 371 Q(ed whether or not the)-.11 E 2.75
(yw)-.165 G(ere ECN-capable.)-2.75 E(Thus, the de)93.6 397 Q
(gree of bene\214t of adding ECN-Capability to SYN/A)-.165 E(CK pack)
-.44 E(ets depends not)-.11 E(only on the o)93.6 410 Q -.165(ve)-.165 G
(rall pack).165 E(et drop rate in the netw)-.11 E(ork, b)-.11 E
(ut also on the queue management)-.22 E
(architecture at the congested link.)93.6 423 Q/F1 11/Times-Bold@0 SF
(6.2.)72 462 Q/F2 13/Times-Bold@0 SF(An Ev)5.5 E(aluation of Differ)-.13
E(ent Responses to ECN-Mark)-.234 E(ed SYN/A)-.13 E(CK P)-.715 E(ack)
-.13 E(ets)-.13 E F0(This document speci\214es that the end-node respon\
ds to the report of an ECN-mark)93.6 488 Q(ed)-.11 E(SYN/A)93.6 501 Q
(CK pack)-.44 E(et by setting the initial congestion windo)-.11 E 2.75
(wt)-.275 G 2.75(oo)-2.75 G(ne se)-2.75 E(gment, instead of its)-.165 E
(possible def)93.6 514 Q(ault v)-.11 E(alue of tw)-.275 E 2.75(ot)-.11 G
2.75(of)-2.75 G(our se)-2.75 E 2.75(gments. W)-.165 F 2.75(ec)-.88 G
(all this ECN+ with NoW)-2.75 E(aiting.)-.88 E(Ho)93.6 527 Q(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(nS).44 G
(ection 3 discussed another possible response to an ECN-mark)-2.75 E
(ed SYN/A)-.11 E(CK)-.44 E(pack)93.6 540 Q(et, of the end-node w)-.11 E
(aiting an R)-.11 E(TT before sending a data pack)-.66 E 2.75(et. W)-.11
F 2.75(ec)-.88 G(all this approach)-2.75 E(ECN+ with W)93.6 553 Q
(aiting.)-.88 E(Simulations comparing the performance with Standard ECN\
 \(without ECN-mark)93.6 579 Q(ed)-.11 E(SYN/A)93.6 592 Q(CK pack)-.44 E
(ets\), ECN+ with NoW)-.11 E(aiting, and ECN+ with W)-.88 E(aiting sho)
-.88 E 2.75(wl)-.275 G(ittle dif)-2.75 E(ference,)-.275 E
(in terms of aggre)93.6 605 Q -.055(ga)-.165 G
(te congestion, between ECN+ with NoW).055 E(aiting and ECN+ with W)-.88
E(aiting.)-.88 E(The details are gi)93.6 618 Q -.165(ve)-.275 G 2.75(ni)
.165 G 2.75(nA)-2.75 G(ppendix A belo)-2.75 E 4.18 -.715(w. O)-.275 H
(ur conclusions are that ECN+ with NoW).715 E(aiting)-.88 E(is perfectl\
y safe, and there are no congestion-related reasons for preferring ECN+\
 with)93.6 631 Q -.88(Wa)93.6 644 S(iting o).88 E -.165(ve)-.165 G 2.75
(rE).165 G(CN+ with NoW)-2.75 E 2.75(aiting. That)-.88 F
(is, there is no need for the TCP end-node to w)2.75 E(ait)-.11 E 2.75
(ar)93.6 657 S(ound-trip time before sending a data pack)-2.75 E
(et after recei)-.11 E(ving an ackno)-.275 E(wledgement of an)-.275 E
(ECN-mark)93.6 670 Q(ed SYN/A)-.11 E(CK pack)-.44 E(et.)-.11 E(Flo)72
733 Q 361.391(yd [P)-.11 F(age 12])-.165 E 0 Cg EP
%%Page: 13 13
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E/F1 11/Times-Bold@0 SF(7.)72 85 Q/F2 14/Times-Bold@0 SF
(Security Considerations)5.5 E F0(TCP pack)93.6 111 Q
(ets carrying the ECT codepoint in IP headers can be mark)-.11 E
(ed rather than dropped)-.11 E(by ECN-capable routers. This raises se)
93.6 124 Q -.165(ve)-.275 G(ral security concerns that we discuss belo)
.165 E -.715(w.)-.275 G("Bad" routers or middlebox)93.6 150 Q(es:)-.165
E(There is a small b)93.6 163 Q
(ut decreasing number of routers or middlebox)-.22 E
(es that drop or reset SYN and)-.165 E(SYN/A)93.6 176 Q(CK pack)-.44 E(\
ets based on the ECN-related \215ags in the TCP header [MAF05], [RFC336\
0].)-.11 E(While there is no e)93.6 189 Q(vidence that an)-.275 E 2.75
(ym)-.165 G(iddlebox)-2.75 E(es drop SYN/A)-.165 E(CK pack)-.44 E
(ets that contain an)-.11 E
(ECN-Capable or CE codepoint in the *IP header*, such beha)93.6 202 Q
(vior cannot be e)-.22 E 2.75(xcluded. Thus,)-.165 F
(as speci\214ed in Section 2, if a SYN/A)93.6 215 Q(CK pack)-.44 E
(et with the ECT or CE codepoint is dropped,)-.11 E
(the TCP node SHOULD resend the SYN/A)93.6 228 Q(CK pack)-.44 E
(et without the ECN-Capable codepoint.)-.11 E(Congestion collapse:)93.6
254 Q(Because TCP SYN/A)93.6 267 Q(CK pack)-.44 E
(ets carrying an ECT codepoint could be ECN-mark)-.11 E(ed instead)-.11
E(of dropped at an ECN-capable router)93.6 280 Q 2.75(,t)-.44 G
(he concern is whether this can either in)-2.75 E -.22(vo)-.44 G -.11
(ke).22 G(congestion, or w)93.6 293 Q
(orsen performance in highly congested scenarios.)-.11 E(Ho)5.5 E(we)
-.275 E -.165(ve)-.275 G .88 -.44(r, a).165 H(fter learning).44 E
(that a SYN/A)93.6 306 Q(CK pack)-.44 E(et w)-.11 E(as ECN-mark)-.11 E
(ed, the sender of that pack)-.11 E(et will only send one data)-.11 E
(pack)93.6 319 Q(et; if this data pack)-.11 E(et is ECN-mark)-.11 E
(ed, the sender will then w)-.11 E(ait for a retransmission)-.11 E 2.75
(timeout. In)93.6 332 R
(addition, routers are free to drop rather than mark arri)2.75 E
(ving pack)-.275 E(ets in times of)-.11 E(high congestion, re)93.6 345 Q
-.055(ga)-.165 G(rdless of whether the pack).055 E(ets are ECN-capable.)
-.11 E(When congestion is)5.5 E -.165(ve)93.6 358 S
(ry high and a router').165 E 2.75(sb)-.605 G(uf)-2.97 E
(fer is full, the router has no choice b)-.275 E
(ut to drop rather than to mark)-.22 E(an arri)93.6 371 Q(ving pack)
-.275 E(et.)-.11 E(The simulations reported in Appendix A sho)93.6 397 Q
2.75(wt)-.275 G(hat e)-2.75 E -.165(ve)-.275 G 2.75(nw).165 G
(ith demanding traf)-2.75 E(\214c mix)-.275 E(es)-.165 E
(dominated by short \215o)93.6 410 Q(ws and high le)-.275 E -.165(ve)
-.275 G(ls of congestion, the aggre).165 E -.055(ga)-.165 G(te pack).055
E(et dropping rates)-.11 E(are not signi\214cantly dif)93.6 423 Q
(ferent with Standard ECN, ECN+ with NoW)-.275 E(aiting, or ECN+ with)
-.88 E -.88(Wa)93.6 436 S 2.75(iting. In).88 F(particular)2.75 E 2.75
(,t)-.44 G(he simulations sho)-2.75 E 2.75(wt)-.275 G
(hat in periods of v)-2.75 E(ery high congestion the)-.165 E(pack)93.6
449 Q(et-marking rate is lo)-.11 E 2.75(ww)-.275 G
(ith or without ECN+, and the use of ECN+ does not)-2.75 E
(signi\214cantly increase the number of dropped or mark)93.6 462 Q
(ed pack)-.11 E(ets.)-.11 E(The simulations sho)93.6 488 Q 2.75(wt)-.275
G(hat ECN+ is most ef)-2.75 E(fecti)-.275 E .33 -.165(ve i)-.275 H 2.75
(nt).165 G(imes of moderate congestion.)-2.75 E(In these)5.5 E(moderate\
-congested scenarios, the use of ECN+ increases the number of ECN-mark)
93.6 501 Q(ed)-.11 E(pack)93.6 514 Q(ets, because ECN+ allo)-.11 E
(ws SYN/A)-.275 E(CK pack)-.44 E(ets to be ECN-mark)-.11 E 2.75(ed. At)
-.11 F(the same time, in)2.75 E(these times of moderate congestion, the\
 use of ECN+ instead of Standard ECN does not)93.6 527 Q
(signi\214cantly af)93.6 540 Q(fect the o)-.275 E -.165(ve)-.165 G
(rall le).165 E -.165(ve)-.275 G(ls of congestion.).165 E
(The simulations sho)93.6 566 Q 2.75(wt)-.275 G
(hat the use of ECN+ is less ef)-2.75 E(fecti)-.275 E .33 -.165(ve i)
-.275 H 2.75(nt).165 G(imes of high congestion;)-2.75 E(the)5.5 E
(simulations sho)93.6 579 Q 2.75(wt)-.275 G
(hat in times of high congestion more pack)-2.75 E
(ets are dropped instead of)-.11 E(mark)93.6 592 Q
(ed, both with Standard ECN and with ECN+.)-.11 E
(In times of high congestion, the b)5.5 E(uf)-.22 E(fer)-.275 E(can o)
93.6 605 Q -.165(ve)-.165 G(r\215o).165 E 1.43 -.715(w, e)-.275 H -.165
(ve).44 G 2.75(nw).165 G(ith Acti)-2.75 E .33 -.165(ve Q)-.275 H
(ueue Management and ECN; when the b).165 E(uf)-.22 E(fer is full)-.275
E(arri)93.6 618 Q(ving pack)-.275 E(ets are dropped rather than mark)
-.11 E(ed, whether the pack)-.11 E(ets are ECN-capable or)-.11 E 2.75
(not. Thus)93.6 631 R(while ECN+ is less ef)2.75 E(fecti)-.275 E .33
-.165(ve i)-.275 H 2.75(nt).165 G
(imes of high congestion, it still doesn')-2.75 E 2.75(tr)-.198 G
(esult in a)-2.75 E(signi\214cant increase in the le)93.6 644 Q -.165
(ve)-.275 G 2.75(lo).165 G 2.75(fc)-2.75 G 2.75(ongestion. More)-2.75 F
(details are gi)2.75 E -.165(ve)-.275 G 2.75(ni).165 G 2.75(nt)-2.75 G
(he appendix.)-2.75 E(Flo)72 733 Q 361.391(yd [P)-.11 F(age 13])-.165 E
0 Cg EP
%%Page: 14 14
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E/F1 11/Times-Bold@0 SF(8.)72 85 Q/F2 14/Times-Bold@0 SF
(Conclusions)5.5 E F0
(This draft speci\214es a modi\214cation to RFC 3168 to allo)93.6 111 Q
2.75(wT)-.275 G(CP nodes to send SYN/A)-2.75 E(CK)-.44 E(pack)93.6 124 Q
(ets as being ECN-Capable.)-.11 E(Making the SYN/A)5.5 E(CK pack)-.44 E
(et ECN-Capable a)-.11 E -.22(vo)-.22 G(ids the).22 E
(high cost to a TCP transfer when a SYN/A)93.6 137 Q(CK pack)-.44 E
(et is dropped by a congested router)-.11 E 2.75(,b)-.44 G(y)-2.75 E
-.22(avo)93.6 150 S(iding the resulting retransmit timeout.).22 E
(This impro)5.5 E -.165(ve)-.165 G 2.75(st).165 G
(he throughput of short)-2.75 E 2.75(connections. The)93.6 163 R
(sender of the SYN/A)2.75 E(CK pack)-.44 E
(et responds to an ECN mark by reducing its)-.11 E
(initial congestion windo)93.6 176 Q 2.75(wf)-.275 G(rom tw)-2.75 E
(o, three, or four se)-.11 E(gments to one se)-.165 E
(gment, reducing the)-.165 E
(subsequent load from that connection on the netw)93.6 189 Q 2.75
(ork. The)-.11 F(addition of ECN-capability to)2.75 E(SYN/A)93.6 202 Q
(CK pack)-.44 E(ets is particularly bene\214cial in the serv)-.11 E(er)
-.165 E(-to-client direction, where)-.22 E(congestion is more lik)93.6
215 Q(ely to occur)-.11 E 5.5(.I)-.605 G 2.75(nt)-5.5 G
(his case, the initial information pro)-2.75 E(vided by the ECN)-.165 E
(marking in the SYN/A)93.6 228 Q(CK pack)-.44 E(et enables the serv)-.11
E(er to more appropriately adjust the initial)-.165 E
(load it places on the netw)93.6 241 Q(ork.)-.11 E(Future w)93.6 267 Q(\
ork will address the more general question of adding ECN-Capability to \
rele)-.11 E -.275(va)-.275 G(nt).275 E(handshak)93.6 280 Q 2.75(ep)-.11
G(ack)-2.75 E(ets in other protocols that use retransmission-based reli\
ability in their setup)-.11 E(phase \(e.g., SCTP)93.6 293 Q 2.75(,D)
-1.221 G(CCP)-2.75 E 2.75(,H)-1.221 G(IP)-2.75 E 2.75(,a)-1.221 G
(nd the lik)-2.75 E(e\).)-.11 E F1(9.)72 319 Q F2(Ackno)5.5 E
(wledgements)-.14 E F0 1.76 -.88(We t)93.6 345 T(hank Anil Ag).88 E(arw)
-.055 E(al, Mark Allman, W)-.11 E(esle)-.88 E 2.75(yE)-.165 G(ddy)-2.75
E 2.75(,J)-.715 G(anardhan Iyeng)-2.75 E(ar)-.055 E 2.75(,a)-.44 G(nd P)
-2.75 E(asi Sarolahti)-.165 E(for feedback on earlier v)93.6 358 Q
(ersions of this draft.)-.165 E F1(A.)72 384 Q F2(Report on Simulations)
5.5 E F0(This section reports on simulations sho)93.6 410 Q
(wing the costs of adding ECN+ in highly-congested)-.275 E 2.75
(scenarios. This)93.6 423 R
(section also reports on simulations for a comparati)2.75 E .33 -.165
(ve e)-.275 H -.275(va)-.11 G(luation between).275 E(ECN+ with NoW)93.6
436 Q(aiting and ECN+ with W)-.88 E(aiting.)-.88 E
(The simulations are run with a range of \214le-size distrib)93.6 462 Q
2.75(utions. As)-.22 F 2.75(ab)2.75 G(aseline, the)-2.75 E 2.75(yu)-.165
G(se the)-2.75 E(empirical hea)93.6 475 Q(vy-tailed distrib)-.22 E
(ution reported in [SCJO01], with a mean \214le size of around 7)-.22 E
2.75(KBytes. This)93.6 488 R<8d6f>2.75 E(w-size distrib)-.275 E
(ution is manipulated by sk)-.22 E -.275(ew)-.11 G(ing the \215o).275 E
2.75(ws)-.275 G(izes to)-2.75 E -.11(wa)-.275 G(rds lo).11 E(wer)-.275 E
(and higher v)93.6 501 Q(alues to get distrib)-.275 E
(utions with mean \214le sizes of 3 KBytes, 5 KBytes, 14 KBytes)-.22 E
(and 17 KBytes.)93.6 514 Q(The congested link is 100 Mbps.)5.5 E
(RED is run in gentle mode, and arri)5.5 E(ving)-.275 E
(ECN-Capable pack)93.6 527 Q(ets are only dropped instead of mark)-.11 E
(ed if the b)-.11 E(uf)-.22 E(fer is full \(and the router)-.275 E
(has no choice\).)93.6 540 Q 1.76 -.88(We e)93.6 566 T(xplore tw).715 E
2.75(oa)-.11 G(lternati)-2.75 E -.165(ve)-.275 G 2.75(sf).165 G
(or a TCP node')-2.75 E 2.75(sr)-.605 G
(esponse to a report of an ECN-mark)-2.75 E(ed)-.11 E(SYN/A)93.6 579 Q
(CK pack)-.44 E 2.75(et. W)-.11 F(ith ECN+ with NoW)-.44 E
(aiting, the TCP node sends a data pack)-.88 E(et)-.11 E
(immediately \(with an initial congestion windo)93.6 592 Q 2.75(wo)-.275
G 2.75(fo)-2.75 G(ne se)-2.75 E 2.75(gment\). W)-.165 F
(ith the alternati)-.44 E .33 -.165(ve E)-.275 H(CN+).165 E(with W)93.6
605 Q(aiting, the TCP node w)-.88 E
(aits a round-trip time before sending a data pack)-.11 E
(et; the sender)-.11 E
(already has one measurement of the round-trip time when the ackno)93.6
618 Q(wledgement for the)-.275 E(SYN/A)93.6 631 Q(CK pack)-.44 E
(et is recei)-.11 E -.165(ve)-.275 G(d.).165 E(In the tables belo)93.6
657 Q 1.43 -.715(w, E)-.275 H(CN+ refers to ECN+ with NoW).715 E
(aiting, where the sender starts)-.88 E(transmitting immediately)93.6
670 Q 2.75(,a)-.715 G(nd ECN+/w)-2.75 E(ait refers to ECN+ with W)-.11 E
(aiting, where the sender)-.88 E -.11(wa)93.6 683 S
(its a round-trip time before sending a data pack).11 E
(et into the netw)-.11 E(ork.)-.11 E(Flo)72 733 Q 361.391(yd [P)-.11 F
(age 14])-.165 E 0 Cg EP
%%Page: 15 15
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(The simulation scripts are a)93.6 85 Q -.275(va)-.22 G
(ilable on [ECN-SYN], along with graphs sho).275 E(wing the)-.275 E
(distrib)93.6 98 Q(ution of response times for the TCP connections.)-.22
E/F1 11/Times-Bold@0 SF(A.1.)72 137 Q/F2 13/Times-Bold@0 SF
(Simulations with RED in P)5.5 E(ack)-.13 E(et Mode)-.13 E F0
(The simulations with RED in pack)93.6 163 Q
(et mode and with the queue in pack)-.11 E(ets sho)-.11 E 2.75(wt)-.275
G(hat ECN+ is)-2.75 E(useful in times of moderate congestion, though it\
 adds little bene\214t in times of high)93.6 176 Q 2.75(congestion. The)
93.6 189 R(simulations sho)2.75 E 2.75(wam)-.275 G
(inimal increase in le)-2.75 E -.165(ve)-.275 G
(ls of congestion with either).165 E(ECN+ with W)93.6 202 Q
(aiting or ECN+ with NoW)-.88 E(aiting, either in terms of pack)-.88 E
(et dropping or marking)-.11 E(rates or in terms of the distrib)93.6 215
Q(ution of responses times.)-.22 E(Thus, the simulations sho)5.5 E 2.75
(wn)-.275 G(o)-2.75 E(problems with ECN+ in times of high congestion, a\
nd no reason to use ECN+ with W)93.6 228 Q(aiting)-.88 E
(instead of ECN+ with NoW)93.6 241 Q(aiting.)-.88 E -.88(Ta)93.6 267 S
(ble 3 sho).88 E(ws the congestion le)-.275 E -.165(ve)-.275 G
(ls for simulations with RED in pack).165 E(et mode, with a queue)-.11 E
(in pack)93.6 280 Q 2.75(ets. T)-.11 F 2.75(oe)-.88 G(xplore a w)-2.915
E(orst-case scenario, these simulations use a traf)-.11 E
(\214c mix with an)-.275 E(unrealistically small \215o)93.6 293 Q 2.75
(ws)-.275 G(ize distrib)-2.75 E(ution, with a mean \215o)-.22 E 2.75(ws)
-.275 G(ize of 3 Kbytes.)-2.75 E -.165(Fo)5.5 G 2.75(re).165 G
(ach table)-2.75 E(sho)93.6 306 Q(wing a particular traf)-.275 E
(\214c load, the three ro)-.275 E(ws sho)-.275 E 2.75(wt)-.275 G
(he number of pack)-2.75 E(ets dropped, the)-.11 E(number of pack)93.6
319 Q(ets ECN-mark)-.11 E(ed, and the aggre)-.11 E -.055(ga)-.165 G
(te pack).055 E(et drop rate, and the three columns)-.11 E(sho)93.6 332
Q 2.75(wt)-.275 G(he simulations with Standard ECN, ECN+ \(NoW)-2.75 E
(aiting\) and ECN+/w)-.88 E(ait.)-.11 E
(The usefulness of ECN+: The \214rst thing to observ)93.6 358 Q 2.75(ei)
-.165 G 2.75(st)-2.75 G(hat for the simulations with the)-2.75 E(some)
93.6 371 Q(what moderate load of 95%, with pack)-.275 E
(et drop rates of 5-6%, the use of ECN+ or)-.11 E(ECN+/w)93.6 384 Q
(ait more than doubled the number of pack)-.11 E(ets mark)-.11 E 2.75
(ed. This)-.11 F(indicates that with)2.75 E(ECN+ or ECN+/w)93.6 397 Q
(ait, man)-.11 E 2.75(yS)-.165 G(YN/A)-2.75 E(CK pack)-.44 E
(ets are mark)-.11 E(ed instead of dropped.)-.11 E
(No increase in congestion: The second thing to observ)93.6 423 Q 2.75
(ei)-.165 G 2.75(st)-2.75 G(hat in all of the simulations, the)-2.75 E
(use of ECN+ or ECN+/w)93.6 436 Q
(ait does not signi\214cantly increase the aggre)-.11 E -.055(ga)-.165 G
(te pack).055 E(et drop rate.)-.11 E(Comparing ECN+ and ECN+/w)93.6 462
Q(ait: The third thing to observ)-.11 E 2.75(ei)-.165 G 2.75(st)-2.75 G
(hat there is little dif)-2.75 E(ference)-.275 E
(between ECN+ and ECN+/w)93.6 475 Q(ait in terms of the aggre)-.11 E
-.055(ga)-.165 G(te pack).055 E(et drop rate.)-.11 E(Thus, there is no)
5.5 E(congestion-related reason to prefer ECN+/w)93.6 488 Q(ait o)-.11 E
-.165(ve)-.165 G 2.75(rE).165 G(CN+.)-2.75 E(Flo)72 733 Q 361.391(yd [P)
-.11 F(age 15])-.165 E 0 Cg EP
%%Page: 16 16
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E -.385(Tr)129.6 85 S(af).385 E(\214c Load = 95%:)-.275 E
19.25(ECN ECN+)168.1 98 R(ECN+/w)13.75 E(ait)-.11 E 11
(------- ------- -------)159.85 111 R 11(Dropped 74,645)129.6 124 R
13.75(64,034 64,983)16.5 F(Mark)129.6 137 Q 16.5(ed 7,639)-.11 F 13.75
(17,681 16,914)16.5 F(Loss rate)129.6 150 Q 16.5(6.05% 5.26% 5.33%)11 F
-.385(Tr)129.6 189 S(af).385 E(\214c Load = 110%:)-.275 E 19.25
(ECN ECN+)168.1 202 R(ECN+/w)13.75 E(ait)-.11 E 11
(------- ------- -------)159.85 215 R 8.25(Dropped 161,644)129.6 228 R
11(163,620 165,196)13.75 F(Mark)129.6 241 Q 16.5(ed 4,375 6,653 6,144)
-.11 F(Loss rate)129.6 254 Q 13.75(10.38% 10.45% 10.53%)8.25 F -.385(Tr)
129.6 293 S(af).385 E(\214c Load = 125%:)-.275 E 19.25(ECN ECN+)168.1
306 R(ECN+/w)13.75 E(ait)-.11 E 11(------- ------- -------)159.85 319 R
8.25(Dropped 257,671)129.6 332 R 11(268,161 264,437)13.75 F(Mark)129.6
345 Q 16.5(ed 2,885 3,712 3,359)-.11 F(Loss rate)129.6 358 Q 13.75
(14.52% 15.00% 14.83%)8.25 F -.385(Tr)129.6 397 S(af).385 E
(\214c Load = 150%:)-.275 E 19.25(ECN ECN+)168.1 410 R(ECN+/w)13.75 E
(ait)-.11 E 11(------- ------- -------)159.85 423 R(Loss rate)129.6 436
Q 13.75(24.36% 24.61% 24.46%)8.25 F -.385(Tr)129.6 475 S(af).385 E
(\214c Load = 200%:)-.275 E 19.25(ECN ECN+)168.1 488 R(ECN+/w)13.75 E
(ait)-.11 E 11(------- ------- -------)159.85 501 R(Loss rate)129.6 514
Q 13.75(29.99% 30.22% 30.23%)8.25 F -.88(Ta)93.6 533 S
(ble 3: Simulations with an a).88 E -.165(ve)-.22 G(rage \215o).165 E
2.75(ws)-.275 G(ize of 3 Kbytes, RED in pack)-2.75 E(et mode, queue in)
-.11 E(pack)93.6 546 Q(ets.)-.11 E/F1 11/Times-Bold@0 SF(A.2.)72 572 Q
/F2 13/Times-Bold@0 SF(Simulations with RED in Byte Mode)5.5 E F0 -.88
(Ta)93.6 598 S(ble 4 belo).88 E 2.75(ws)-.275 G(ho)-2.75 E
(ws simulations with RED in byte mode and the queue in bytes.)-.275 E
(Lik)5.5 E 2.75(et)-.11 G(he)-2.75 E(simulations with RED in pack)93.6
611 Q(et mode, there is no signi\214cant increase in aggre)-.11 E -.055
(ga)-.165 G(te).055 E(congestion with the use of ECN+ or ECN+/w)93.6 624
Q(ait, and no congestion-related reason to prefer)-.11 E(ECN+/w)93.6 637
Q(ait o)-.11 E -.165(ve)-.165 G 2.75(rE).165 G(CN+.)-2.75 E(Ho)93.6 663
Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, u).165 H(nlik).44 E 2.75(et)
-.11 G(he simulations with RED in pack)-2.75 E
(et mode, the simulations with RED in)-.11 E(byte mode sho)93.6 676 Q
2.75(wl)-.275 G(ittle bene\214t from the use of ECN+ or ECN+/w)-2.75 E
(ait, in that the pack)-.11 E(et)-.11 E
(marking rate with ECN+ or ECN+/w)93.6 689 Q(ait is not much dif)-.11 E
(ferent than the pack)-.275 E(et marking rate)-.11 E(Flo)72 733 Q
361.391(yd [P)-.11 F(age 16])-.165 E 0 Cg EP
%%Page: 17 17
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(with Standard ECN.)93.6 85 Q
(This is because with RED in byte mode, small pack)5.5 E(ets lik)-.11 E
2.75(eS)-.11 G(YN/A)-2.75 E(CK)-.44 E(pack)93.6 98 Q
(ets are rarely dropped or mark)-.11 E(ed - that is, there is no dra)
-.11 E(wback from the use of ECN+ in)-.165 E(these scenarios, b)93.6 111
Q(ut not much need for ECN+ either)-.22 E 2.75(,i)-.44 G 2.75(nas)-2.75
G(cenario where small pack)-2.75 E(ets are)-.11 E(unlik)93.6 124 Q
(ely to be dropped or mark)-.11 E(ed.)-.11 E -.385(Tr)129.6 156 S(af)
.385 E(\214c Load = 95%:)-.275 E 19.25(ECN ECN+)168.1 169 R(ECN+/w)13.75
E(ait)-.11 E 11(------- ------- -------)159.85 182 R 11(Dropped 13,044)
129.6 195 R 13.75(13,323 14,855)16.5 F(Mark)129.6 208 Q 13.75
(ed 18,880 19,175 19,049)-.11 F(Loss rate)129.6 221 Q 16.5
(1.13% 1.16% 1.29%)11 F -.385(Tr)129.6 260 S(af).385 E
(\214c Load = 110%:)-.275 E 19.25(ECN ECN+)168.1 273 R(ECN+/w)13.75 E
(ait)-.11 E 11(------- ------- -------)159.85 286 R 11(Dropped 84,809)
129.6 299 R 13.75(83,013 83,564)16.5 F(Mark)129.6 312 Q 16.5
(ed 4,086 4,644 4,826)-.11 F(Loss rate)129.6 325 Q 16.5
(5.90% 5.78% 5.81%)11 F -.385(Tr)129.6 364 S(af).385 E
(\214c Load = 125%:)-.275 E 19.25(ECN ECN+)168.1 377 R(ECN+/w)13.75 E
(ait)-.11 E 11(------- ------- -------)159.85 390 R 8.25
(Dropped 157,305)129.6 403 R 11(157,435 158,368)13.75 F(Mark)129.6 416 Q
16.5(ed 2,183 2,363 2,663)-.11 F(Loss rate)129.6 429 Q 16.5
(9.89% 9.87% 9.93%)11 F -.88(Ta)93.6 448 S(ble 4: Simulations with an a)
.88 E -.165(ve)-.22 G(rage \215o).165 E 2.75(ws)-.275 G
(ize of 3 Kbytes, RED in byte mode, queue in)-2.75 E(bytes.)93.6 461 Q
/F1 14/Times-Bold@0 SF(Normati)72 487 Q .28 -.14(ve R)-.14 H(efer).14 E
(ences)-.252 E F0([RFC 2119] S. Bradner)93.6 513 Q 2.75(,K)-.44 G .33
-.165(ey w)-3.025 H(ords for use in RFCs to Indicate Requirement Le).055
E -.165(ve)-.275 G(ls, RFC).165 E(2119, March 1997.)93.6 526 Q
([RFC3168] K.K. Ramakrishnan, S. Flo)93.6 552 Q
(yd, and D. Black, The Addition of Explicit)-.11 E
(Congestion Noti\214cation \(ECN\) to IP)93.6 565 Q 2.75(,R)-1.221 G
(FC 3168, Proposed Standard, September 2001.)-2.75 E F1(Inf)72 591 Q
(ormati)-.35 E .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252 E F0
([ECN+] A. K)93.6 617 Q(uzmano)-.165 E(vic, The Po)-.165 E
(wer of Explicit Congestion Noti\214cation, SIGCOMM 2005.)-.275 E
([ECN-SYN] ECN-SYN web page with simulation scripts, URL to be added.)
93.6 643 Q([MAF05] A. Medina, M. Allman, and S. Flo)93.6 669 Q 2.75
(yd. Measuring)-.11 F(the Ev)2.75 E(olution of T)-.22 E(ransport)-.385 E
(Protocols in the Internet, A)93.6 682 Q(CM CCR, April 2005.)-.44 E(Flo)
72 733 Q 361.391(yd [P)-.11 F(age 17])-.165 E 0 Cg EP
%%Page: 18 18
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E([PI] C. Hollot, V)93.6 85 Q 2.75(.M)-1.419 G(isra, W)-2.75
E 2.75(.G)-1.012 G(ong, and D. T)-2.75 E -.275(ow)-.88 G(sle).275 E 1.43
-.715(y, O)-.165 H 2.75(nD).715 G(esigning Impro)-2.75 E -.165(ve)-.165
G 2.75(dC).165 G(ontrollers for)-2.75 E -.605(AQ)93.6 98 S 2.75(MR).605
G(outers Supporting TCP Flo)-2.75 E(ws, April 1998.)-.275 E([RED] Flo)
93.6 124 Q(yd, S., and Jacobson, V)-.11 E 5.5(.R)-1.419 G
(andom Early Detection g)-5.5 E(ate)-.055 E -.11(wa)-.275 G
(ys for Congestion).11 E -1.342 -.814(Av o)93.6 137 T(idance .).814 E
(IEEE/A)5.5 E(CM T)-.44 E(ransactions on Netw)-.385 E(orking, V)-.11 E
(.1 N.4, August 1993.)-1.419 E([REM] S. Athuraliya, V)93.6 163 Q 2.75
(.H)-1.419 G 2.75(.L)-2.75 G(i, S. H. Lo)-2.75 E 2.75(wa)-.275 G
(nd Q. Y)-2.75 E(in, REM: Acti)-.605 E .33 -.165(ve Q)-.275 H
(ueue Management,).165 E(IEEE Netw)93.6 176 Q(ork, May 2001.)-.11 E([RF\
C2309] B. Braden et al., Recommendations on Queue Management and Conges\
tion)93.6 202 Q -1.342 -.814(Av o)93.6 215 T
(idance in the Internet, RFC 2309, April 1998.).814 E
([RFC2581] M. Allman, V)93.6 241 Q 2.75(.P)-1.419 G(axson, and W)-2.915
E 2.75(.S)-1.012 G(te)-2.75 E -.165(ve)-.275 G
(ns, TCP Congestion Control, RFC 2581,).165 E(April 1999.)93.6 254 Q
([RFC2988] V)93.6 280 Q 2.75(.P)-1.419 G
(axson and M. Allman, Computing TCP')-2.915 E 2.75(sR)-.605 G
(etransmission T)-2.75 E(imer)-.385 E 2.75(,R)-.44 G(FC 2988,)-2.75 E
(No)93.6 293 Q -.165(ve)-.165 G(mber 2000.).165 E
([RFC3042] M. Allman, H. Balakrishnan, and S. Flo)93.6 319 Q
(yd, Enhancing TCP')-.11 E 2.75(sL)-.605 G(oss Reco)-2.75 E -.165(ve)
-.165 G(ry).165 E(Using Limited T)93.6 332 Q
(ransmit, RFC 3042, Proposed Standard, January 2001.)-.385 E
([RFC3360] S. Flo)93.6 358 Q
(yd, Inappropriate TCP Resets Considered Harmful, RFC 3360, August)-.11
E(2002.)93.6 371 Q([RFC3390] M. Allman, S. Flo)93.6 397 Q(yd, and C. P)
-.11 E(artridge, Increasing TCP')-.165 E 2.75(sI)-.605 G(nitial W)-2.75
E(indo)-.44 E 1.43 -.715(w, R)-.275 H(FC).715 E(3390, October 2002.)93.6
410 Q([SCJO01] F)93.6 436 Q 2.75(.S)-.88 G(mith, F)-2.75 E 2.75(.C)-.88
G(ampos, K. Jef)-2.75 E -.11(fa)-.275 G 1.43 -.715(y, D).11 H 2.75(.O)
.715 G(tt, What {TCP/IP} Protocol Headers Can T)-2.75 E(ell)-.77 E
(us about the W)93.6 449 Q(eb, SIGMETRICS, June 2001.)-.88 E 5.5
([SYN-COOK] Dan)93.6 475 R(J. Bernstein, SYN cookies, 1997, see also)
2.75 E(<http://cr)93.6 488 Q(.yp.to/syncookies.html>)-.605 E([T)93.6 514
Q(ools] S. Flo)-.88 E(yd and E. K)-.11 E(ohler)-.385 E 2.75(,T)-.44 G
(ools for the Ev)-3.63 E(aluation of Simulation and T)-.275 E(estbed)
-.77 E(Scenarios, Internet-draft draft-irtf-tmr)93.6 527 Q
(g-tools-03, w)-.198 E(ork in progress, December 2006.)-.11 E/F1 14
/Times-Bold@0 SF(IAN)72 553 Q 3.5(AC)-.28 G(onsiderations)-3.5 E F0
(There are no IAN)93.6 579 Q 2.75(Ac)-.385 G(onsiderations re)-2.75 E
-.055(ga)-.165 G(rding this document.).055 E F1 -.7(AU)72 618 S
(THORS' ADDRESSES).7 E F0(Flo)72 733 Q 361.391(yd [P)-.11 F(age 18])
-.165 E 0 Cg EP
%%Page: 19 19
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E/F1 11/Courier@0 SF(Aleksandar Kuzmanovic)93.6 85 Q
(Phone: +1 \(847\) 467-5519)93.6 98 Q(Northwestern University)93.6 111 Q
(Email: akuzma at northwestern.edu)93.6 124 Q
(URL: http://cs.northwestern.edu/~a)93.6 137 Q(Amit Mondal)93.6 163 Q
(Northwestern University)93.6 176 Q(Email: a-mondal at northwestern.edu)
93.6 189 Q(Sally Floyd)93.6 215 Q(Phone: +1 \(510\) 666-2989)93.6 228 Q
(ICIR \(ICSI Center for Internet Research\))93.6 241 Q
(Email: floyd at icir.org)93.6 254 Q(URL: http://www.icir.org/floyd/)
93.6 267 Q(K. K. Ramakrishnan)93.6 293 Q(Phone: +1 \(973\) 360-8764)93.6
306 Q(AT&T Labs Research)93.6 319 Q(Email: kkrama at research.att.com)
93.6 332 Q(URL: http://www.research.att.com/info/kkrama)93.6 345 Q/F2 14
/Times-Bold@0 SF(Full Copyright Statement)72 386.1 Q F0(Cop)93.6 402.7 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E(This document is\
 subject to the rights, licenses and restrictions contained in BCP 78, \
and)93.6 419.3 Q -.165(ex)93.6 432.3 S
(cept as set forth therein, the authors retain all their rights.).165 E
(This document and the information contained herein are pro)93.6 448.9 Q
(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)93.6 461.9 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)93.6 474.9 Q 2.75(,T)
-1.419 G(HE IETF TR)-2.75 E(UST AND THE)-.44 E(INTERNET ENGINEERING T)
93.6 487.9 Q(ASK FORCE DISCLAIM ALL W)-1.023 E(ARRANTIES, EXPRESS)-1.32
E(OR IMPLIED, INCLUDING B)93.6 500.9 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)-2.75 E(USE OF THE INFORMA)93.6 513.9 Q
(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G(NFRINGE ANY RIGHTS OR ANY)
-2.75 E(IMPLIED W)93.6 526.9 Q(ARRANTIES OF MERCHANT)-1.32 E
(ABILITY OR FITNESS FOR A)-1.023 E -1.012(PA)93.6 539.9 S -.66(RT)1.012
G(ICULAR PURPOSE.).66 E F2(Intellectual Pr)72 565.9 Q(operty)-.252 E F0
(The IETF tak)93.6 582.5 Q(es no position re)-.11 E -.055(ga)-.165 G
(rding the v).055 E(alidity or scope of an)-.275 E 2.75(yI)-.165 G
(ntellectual Property)-2.75 E(Rights or other rights that might be clai\
med to pertain to the implementation or use of the)93.6 595.5 Q
(technology described in this document or the e)93.6 608.5 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)93.6 621.5 Q -.275(va)-.22 G
(ilable; nor does it represent that it has made an).275 E 2.75(yi)-.165
G(ndependent ef)-2.75 E(fort)-.275 E(to identify an)93.6 634.5 Q 2.75
(ys)-.165 G(uch rights.)-2.75 E
(Information on the procedures with respect to rights in RFC)5.5 E
(documents can be found in BCP 78 and BCP 79.)93.6 647.5 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)93.6
673.5 Q 2.75(ya)-.165 G(ssurances of licenses to be)-2.75 E(made a)93.6
686.5 Q -.275(va)-.22 G(ilable, or the result of an attempt made to obt\
ain a general license or permission for).275 E(Flo)72 733 Q 361.391
(yd [P)-.11 F(age 19])-.165 E 0 Cg EP
%%Page: 20 20
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 46.094(-DRAFT ECN)-1.012 F
(AND SYN/A)2.75 E(CK P)-.44 E -.44(AC)-1.012 G 93.658(KETS June).44 F
(2007)2.75 E(the use of such proprietary rights by implementers or user\
s of this speci\214cation can be)93.6 85 Q
(obtained from the IETF on-line IPR repository at http://www)93.6 98 Q
(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E(The IETF in)93.6 124 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)93.6 137 Q
-.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)93.6 150 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
93.6 163 Q(g.)-.198 E(Flo)72 733 Q 361.391(yd [P)-.11 F(age 20])-.165 E
0 Cg EP
%%Trailer
end
%%EOF