One document matched: draft-floyd-tsvwg-besteffort-00.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.19.1
%%CreationDate: Fri Jun 29 15:22:18 2007
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.19 1
%%Pages: 13
%%PageOrder: Ascend
%%DocumentMedia: Default 595 842 0 () ()
%%Orientation: Portrait
%%EndComments
%%BeginDefaults
%%PageMedia: Default
%%EndDefaults
%%BeginProlog
%%BeginResource: procset grops 1.19 1
%!PS-Adobe-3.0 Resource-ProcSet
/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
/setcmykcolor where{
pop
/Fk{
setcmykcolor fill
}bind def
}if
/Fg{
setgray fill
}bind def
/FL/fill load def
/LW/setlinewidth load def
/Cr/setrgbcolor load def
/setcmykcolor where{
pop
/Ck/setcmykcolor load def
}if
/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
/setpagedevice{}def
}bind def
/PEND{
clear
countdictstack CNT sub{end}repeat
level1 restore
}bind def
end def
/setpacking where{
pop
setpacking
}if
%%EndResource
%%EndProlog
%%BeginSetup
%%BeginFeature: *PageSize Default
<< /PageSize [ 595 842 ] /ImagingBBox null >> setpagedevice
%%EndFeature
%%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
%%EndSetup
%%Page: 1 1
%%BeginPageSetup
BP
%%EndPageSetup
/F0 10/Courier-Bold@0 SF(Internet Engineering Task Force)72 98 Q
(S. Floyd)198 E 288(INTERNET-DRAFT M.)72 111 R(Allman)6 E
(Intended status: Informational)72 124 Q(ICIR)228 E
(Expires: 29 December 2007)72 137 Q(29 June 2007)210 E/F1 14
/Times-Bold@0 SF(Comments on the Usefulness of Simple Best-Eff)115.464
162 Q(ort T)-.35 E(raf\214c)-1.036 E(draft-\215oyd-tsvwg-besteff)187.669
175 Q(ort-00.ps)-.35 E/F2 11/Times-Bold@0 SF(Status of this Memo)72 207
Q/F3 11/Times-Roman@0 SF
(By submitting this Internet-Draft, each author represents that an)93.6
223.6 Q 2.75(ya)-.165 G(pplicable patent or other)-2.75 E
(IPR claims of which he or she is a)93.6 236.6 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)93.6 249.6
Q -.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)93.6 266.2 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)93.6 279.2 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.)93.6 292.2 Q
(Internet-Drafts are draft documents v)93.6 308.8 Q
(alid for a maximum of six months and may be updated,)-.275 E
(replaced, or obsoleted by other documents at an)93.6 321.8 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
334.8 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)93.6
351.4 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)93.6 364.4 Q
(The list of Internet-Draft Shado)93.6 381 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)93.6 394 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)93.6 410.6 Q(xpire on 29 December 2007.)
-.165 E F2(Copyright Notice)72 436.6 Q F3(Cop)93.6 453.2 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E F2(Abstract)72
479.2 Q F3(This document presents some observ)93.6 495.8 Q
(ations on "simple best-ef)-.275 E(fort" traf)-.275 E
(\214c, de\214ned loosely for)-.275 E
(the purposes of this document as Internet traf)93.6 508.8 Q
(\214c that is not co)-.275 E -.165(ve)-.165 G
(red by Quality of Service).165 E
(mechanisms, congestion-based pricing, or the lik)93.6 521.8 Q 2.75
(e. One)-.11 F(observ)2.75 E(ation is that simple best-ef)-.275 E(fort)
-.275 E(traf)93.6 534.8 Q(\214c serv)-.275 E
(es a useful role in the Internet, and is w)-.165 E(orth k)-.11 E 2.75
(eeping. While)-.11 F(traf)2.75 E(\214c with Quality of)-.275 E
(Service mechanisms, congestion-based pricing, or the lik)93.6 547.8 Q
2.75(ec)-.11 G(an also be useful, we belie)-2.75 E .33 -.165(ve t)-.275
H(hat).165 E(the)93.6 560.8 Q 2.75(ya)-.165 G
(re useful as **adjuncts** to simple best-ef)-2.75 E(fort traf)-.275 E
(\214c, not as **replacements** of)-.275 E(simple best-ef)93.6 573.8 Q
(fort traf)-.275 E 2.75(\214c. A)-.275 F(second observ)2.75 E
(ation is that for simple best-ef)-.275 E(fort traf)-.275 E(\214c, some)
-.275 E(form of rough \215o)93.6 586.8 Q 2.75(wr)-.275 G(ate f)-2.75 E
(airness is a useful goal for resource allocation, where "\215o)-.11 E
2.75(wr)-.275 G(ate)-2.75 E -.11(fa)93.6 599.8 S
(irness" is de\214ned by the goal of equal \215o).11 E 2.75(wr)-.275 G
(ates for dif)-2.75 E(ferent \215o)-.275 E(ws.)-.275 E F2 -1.012(Ta)72
625.8 S(ble of Contents)1.012 E/F4 10/Times-Roman@0 SF(1. Introduction)
93.6 650.8 Q F3 -2.25(.................................................\
.......................................................................\
 ....)4.46 F F4(2)A(2. On Simple Best-Ef)93.6 662.8 Q(fort T)-.25 E(raf)
-.35 E<8c63>-.25 E F3 -2.25(...........................................\
.................................................... ....)3.24 F F4(3)A
(2.1. The Usefulness of Simple Best-Ef)108.6 674.8 Q(fort T)-.25 E(raf)
-.35 E<8c63>-.25 E F3 -2.25
(................................................................. ....)
2.7 F F4(3)A(2.2. The Limitations of Simple Best-Ef)108.6 686.8 Q
(fort T)-.25 E(raf)-.35 E<8c63>-.25 E F3 -2.25
(................................................................ ....)
3.21 F F4(3)A F3(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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
/F1 10/Times-Roman@0 SF(2.2.1. QoS)133.6 84 Q F0 -2.25(................\
.......................................................................\
......................... ....)3.12 F F1(3)A
(2.2.2. The Enforcement of F)133.6 96 Q(airness)-.15 E F0 -2.25(.......\
..................................................................... .\
...)3.69 F F1(4)A(3. On Flo)93.6 108 Q(w-Rate F)-.25 E
(airness for Simple Best-Ef)-.15 E(fort T)-.25 E(raf)-.35 E<8c63>-.25 E
F0 -2.25
(............................................................ ....)5.18
F F1(5)A(3.1. The Usefulness of Flo)108.6 120 Q(w-Rate F)-.25 E(airness)
-.15 E F0 -2.25(.......................................................\
................... ....)4.99 F F1(5)A(3.2. The Limitations of Flo)108.6
132 Q(w-Rate F)-.25 E(airness)-.15 E F0 -2.25(.........................\
................................................. ....)2.75 F F1(6)A
(4. On the Dif)93.6 144 Q(\214culties of Incremental Deplo)-.25 E(yment)
-.1 E F0 -2.25(........................................................\
.............. ....)4.28 F F1(8)A(5. Related W)93.6 156 Q(ork)-.8 E F0
-2.25(.................................................................\
..................................................... ....)4.38 F F1(9)A
(5.1. From the IETF)108.6 168 Q F0 -2.25(..............................\
.......................................................................\
....... ....)4.4 F F1(9)A(5.2. From Else)108.6 180 Q(where)-.25 E F0
-1.75(.................................................................\
......................................... ..)4.33 F F1(10)A
(6. Security Considerations)93.6 192 Q F0 -1.75(.......................\
.......................................................................\
......... ..)4.82 F F1(10)A(7. IAN)93.6 204 Q 2.5(AC)-.35 G
(onsiderations)-2.5 E F0 -1.75(........................................\
................................................................... ..)
2.51 F F1(10)A(8. Conclusions)93.6 216 Q F0 -1.75(.....................\
.......................................................................\
............................ ..)4.45 F F1(10)A(9. Ackno)93.6 228 Q
(wledgements)-.25 E F0 -1.75(..........................................\
................................................................... ..)
4.97 F F1(10)A(Informati)93.6 240 Q .3 -.15(ve R)-.25 H(eferences).15 E
F0 -1.75(..............................................................\
.............................................. ..)3.17 F F1(10)A
(Full Cop)93.6 252 Q(yright Statement)-.1 E F0 -1.75(..................\
.......................................................................\
................ ..)3.57 F F1(13)A(Intellectual Property)93.6 264 Q F0
-1.75(.................................................................\
............................................... ..)3.97 F F1(13)A/F2 11
/Times-Bold@0 SF(1.)72 293.6 Q/F3 14/Times-Bold@0 SF(Intr)5.5 E
(oduction)-.252 E F0(This document gi)93.6 319.6 Q -.165(ve)-.275 G 2.75
(ss).165 G(ome observ)-2.75 E(ations on the role of simple best-ef)-.275
E(fort traf)-.275 E(\214c in the)-.275 E 2.75(Internet. W)93.6 332.6 R
2.75(ed)-.88 G(e\214ne the term "simple best-ef)-2.75 E(fort traf)-.275
E(\214c" to a)-.275 E -.22(vo)-.22 G(id unproducti).22 E .33 -.165(ve s)
-.275 H(emantic).165 E(discussions about whether the phrase "best-ef)
93.6 345.6 Q(fort traf)-.275 E(\214c" does or does not include traf)
-.275 E(\214c with)-.275 E(cost-based f)93.6 358.6 Q
(airness or some other form of congestion-based pricing.)-.11 E -.165
(Fo)5.5 G 2.75(rt).165 G(he purposes of this)-2.75 E
(document, we de\214ne "simple best-ef)93.6 371.6 Q(fort traf)-.275 E
(\214c" as traf)-.275 E(\214c that does not *rely* on the)-.275 E(*dif)
93.6 384.6 Q(ferential treatment* of \215o)-.275 E
(ws either in routers or in policers, enforcers, or other)-.275 E
(middlebox)93.6 397.6 Q(es along the path.)-.165 E("Simple best-ef)5.5 E
(fort traf)-.275 E(\214c" in the current Internet uses end-to-)-.275 E
(end transport protocols \(e.g., TCP)93.6 410.6 Q 2.75(,U)-1.221 G(DP)
-2.75 E 2.75(,o)-1.221 G 2.75(ro)-2.75 G
(thers\), with minimal requirements within the)-2.75 E(netw)93.6 423.6 Q
(ork in terms of resource allocation.)-.11 E(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, t).165 H(hat is not the only possible).44 E
(implementation of a simple best-ef)93.6 436.6 Q(fort service.)-.275 E
(As an e)5.5 E(xample, "simple best-ef)-.165 E(fort traf)-.275 E(\214c")
-.275 E(could also be implemented in an infrastructure with F)93.6 449.6
Q(air Queueing scheduling in congested)-.165 E 2.75(routers. "Simple)
93.6 462.6 R(best-ef)2.75 E(fort traf)-.275 E
(\214c" is the dominant traf)-.275 E
(\214c class in the current Internet.)-.275 E
(In contrast to "simple best-ef)93.6 488.6 Q(fort traf)-.275 E
(\214c", intserv or dif)-.275 E(fserv-enabled traf)-.275 E
(\214c relies on)-.275 E(dif)93.6 501.6 Q
(ferential scheduling mechanisms at congested routers, with pack)-.275 E
(ets from dif)-.11 E(ferent intserv)-.275 E(or dif)93.6 514.6 Q
(fserv classes recei)-.275 E(ving dif)-.275 E(ferent treatment.)-.275 E
(Similarly)5.5 E 2.75(,i)-.715 G 2.75(nc)-2.75 G
(ontrast to "simple best-ef)-2.75 E(fort)-.275 E(traf)93.6 527.6 Q
(\214c", cost-based f)-.275 E(airness [B07] w)-.11 E(ould most lik)-.11
E(ely require the deplo)-.11 E(yment of traf)-.11 E<8c63>-.275 E(markin\
g \(e.g., ECN\) at congested routers, along with policing mechanisms ne\
ar the tw)93.6 540.6 Q 2.75(oe)-.11 G(nds)-2.75 E(of the connection pro)
93.6 553.6 Q(viding dif)-.165 E(ferential treatment for pack)-.275 E
(ets in dif)-.11 E(ferent \215o)-.275 E(ws or in dif)-.275 E(ferent)
-.275 E(traf)93.6 566.6 Q(\214c classes.)-.275 E(Intserv/dif)5.5 E
(fserv)-.275 E 2.75(,c)-.715 G(ost-based f)-2.75 E
(airness, and congestion-based pricing w)-.11 E(ould also)-.11 E
(require comple)93.6 579.6 Q 2.75(xe)-.165 G
(conomic relationships among Internet Service Pro)-2.75 E
(viders \(ISPs\), and)-.165 E(between end-users and ISPs.)93.6 592.6 Q(\
This document suggests that it is important to retain the class of "sim\
ple best-ef)93.6 618.6 Q(fort traf)-.275 E(\214c")-.275 E
(\(though hopefully augmented by a wider deplo)93.6 631.6 Q
(yment of other classes of service\).)-.11 E(Further)5.5 E(,)-.44 E
(this document suggests that some form of rough \215o)93.6 644.6 Q
(w-rate f)-.275 E(airness is a perfectly appropriate)-.11 E
(goal for simple best-ef)93.6 657.6 Q(fort traf)-.275 E 2.75(\214c. W)
-.275 F 2.75(ed)-.88 G 2.75(on)-2.75 G(ot ar)-2.75 E
(gue in this document that \215o)-.198 E(w-rate f)-.275 E(airness is)
-.11 E(the *only possible* or *only desirable* resource allocation goal\
 for simple best-ef)93.6 670.6 Q(fort traf)-.275 E(\214c.)-.275 E 1.76
-.88(We w)93.6 683.6 T(ould maintain, ho).77 E(we)-.275 E -.165(ve)-.275
G .88 -.44(r, t).165 H
(hat it is an appropriate resource allocation goal for simple).44 E
(best-ef)93.6 696.6 Q(fort traf)-.275 E
(\214c in the current Internet, e)-.275 E -.22(vo)-.275 G
(lving from the Internet').22 E 2.75(sp)-.605 G(ast of TCP and UDP)-2.75
E(.)-1.221 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
(This document w)93.6 85 Q(as moti)-.11 E -.275(va)-.275 G
(ted by [B07], an internet-draft on "Flo).275 E 2.75(wR)-.275 G(ate F)
-2.75 E(airness:)-.165 E(Dismantling a Religion" that asserts in the ab\
stract that "Comparing \215o)93.6 98 Q 2.75(wr)-.275 G(ates should ne)
-2.75 E -.165(ve)-.275 G(r).165 E(ag)93.6 111 Q
(ain be used for claims of f)-.055 E(airness in production netw)-.11 E
2.75(orks." This)-.11 F(document does not)2.75 E(attempt to be a reb)
93.6 124 Q(uttal to [B07], or to answer an)-.22 E 2.75(yo)-.165 G 2.75
(ra)-2.75 G(ll of the issues raised in [B07], or to)-2.75 E(gi)93.6 137
Q .33 -.165(ve t)-.275 H(he "intellectual heritage" for \215o).165 E
(w-based f)-.275 E(airness in philosoph)-.11 E 2.75(yo)-.055 G 2.75(rs)
-2.75 G(ocial science, or to)-2.75 E
(commit the authors of this document to an e)93.6 150 Q
(xtended dialogue with the author of [B07].)-.165 E(This)5.5 E
(document is simply a separate vie)93.6 163 Q
(wpoint on some related topics.)-.275 E/F1 11/Times-Bold@0 SF(2.)72 189
Q/F2 14/Times-Bold@0 SF(On Simple Best-Eff)5.5 E(ort T)-.35 E(raf\214c)
-1.036 E F0(This section mak)93.6 215 Q(es some observ)-.11 E
(ations on the usefulness and limitations of the class of)-.275 E
(simple best-ef)93.6 228 Q(fort traf)-.275 E
(\214c, in comparison with QoS-enabled traf)-.275 E(\214c, traf)-.275 E
(\214c with cost-based)-.275 E -.11(fa)93.6 241 S
(irness or congestion-based pricing, and the lik).11 E(e.)-.11 E F1
(2.1.)72 267 Q/F3 13/Times-Bold@0 SF(The Usefulness of Simple Best-Eff)
5.5 E(ort T)-.325 E(raf\214c)-.962 E F0
(This section discusses some of the useful aspects of simple best-ef)
93.6 293 Q(fort traf)-.275 E(\214c.)-.275 E
(Minimal technical demands on the netw)93.6 319 Q(ork infrastructure:)
-.11 E(Simple best-ef)115.2 335.6 Q(fort traf)-.275 E
(\214c, as implemented in the current Internet, mak)-.275 E(es minimal)
-.11 E(technical demands on the infrastructure.)115.2 348.6 Q
(There are no technical requirements for)5.5 E(scheduling mechanisms an\
d queue management mechanisms in routers, or for)115.2 361.6 Q
(enforcement mechanisms along the path.)115.2 374.6 Q
(This is a source of man)5.5 E 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he weaknesses as)-2.75 E(well as man)115.2 387.6 Q 2.75(yo)-.165 G 2.75
(ft)-2.75 G(he strengths of simple best-ef)-2.75 E(fort traf)-.275 E
(\214c.)-.275 E(Minimal demands in terms of economic infrastructure:)
93.6 404.2 Q(Simple best-ef)115.2 420.8 Q(fort traf)-.275 E(\214c mak)
-.275 E(es minimal demands in terms of economic infrastructure,)-.11 E
(relying on f)115.2 433.8 Q(airly simple pair)-.11 E
(-wise economic relationships among ISPs, and between)-.22 E
(users and their immediate ISPs.)115.2 446.8 Q(Usefulness in the real w)
93.6 463.4 Q(orld:)-.11 E(The Internet is chugging along, ho)115.2 480 Q
(we)-.275 E -.165(ve)-.275 G 2.75(ri).165 G(mperfectly)-2.75 E 5.5(.A)
-.715 G 2.75(sd)-5.5 G(iscussed belo)-2.75 E 1.43 -.715(w, s)-.275 H
(imple best-).715 E(ef)115.2 493 Q(fort traf)-.275 E
(\214c is not optimal.)-.275 E(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88
-.44(r, e).165 H(xperience in the Internet has sho).275 E
(wn that there is)-.275 E -.275(va)115.2 506 S(lue in ha).275 E
(ving a mechanism that generally allo)-.22 E
(ws all users to get a portion of the)-.275 E
(resources, while still pre)115.2 519 Q -.165(ve)-.275 G
(nting congestion collapse.).165 E F1(2.2.)72 548.6 Q F3
(The Limitations of Simple Best-Eff)5.5 E(ort T)-.325 E(raf\214c)-.962 E
F0(This section discusses some of the limitations of simple best-ef)93.6
574.6 Q(fort traf)-.275 E(\214c.)-.275 E F1 2.75(2.2.1. QoS)72 613.6 R
F0(Some users w)93.6 639.6 Q(ould be happ)-.11 E 2.75(yt)-.11 G 2.75(op)
-2.75 G(ay for more bandwidth, less delay)-2.75 E 2.75(,l)-.715 G
(ess jitter)-2.75 E 2.75(,o)-.44 G 2.75(rf)-2.75 G -.275(ew)-2.75 G
(er pack).275 E(et)-.11 E 2.75(drops. It)93.6 652.6 R(is desirable to a\
ccommodate such goals in the Internet architecture while)2.75 E
(preserving a suf)93.6 665.6 Q
(\214cient amount of bandwidth for best-ef)-.275 E(fort traf)-.275 E
(\214c.)-.275 E(One of the ob)93.6 691.6 Q(vious dangers of simple impl\
ementations of QoS mechanisms or of cost-based)-.165 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
-.11(fa)93.6 85 S
(irness, in the absence of the protection of simple best-ef).11 E
(fort traf)-.275 E(\214c, w)-.275 E(ould be that the users)-.11 E
(with more mone)93.6 98 Q 2.75(y*)-.165 G
(could* end up completely shutting out users with less mone)-2.75 E 2.75
(yi)-.165 G 2.75(nt)-2.75 G(imes of)-2.75 E 2.75(congestion. There)93.6
111 R(seems to be f)2.75 E(airly widespread agreement that this w)-.11 E
(ould not be a)-.11 E(desirable goal.)93.6 124 Q
(As a sample of the range of positions, the Internet Society')93.6 163 Q
2.75(sI)-.605 G(nternet 2020 Initiati)-2.75 E -.165(ve)-.275 G 2.75(,e)
.165 G(ntitled)-2.75 E("The Internet is \(still\) for Ev)93.6 176 Q
(eryone", states that "we remain committed to the openness that)-.165 E
(ensures equal access and full participation for e)93.6 189 Q -.165(ve)
-.275 G(ry user" [Internet2020].).165 E
(The wide-ranging discussion of "netw)93.6 215 Q
(ork neutrality" in the United States includes adv)-.11 E(ocates)-.22 E
(of se)93.6 228 Q -.165(ve)-.275 G(ral dif).165 E(ferent positions, inc\
luding that of "absolute non-discrimination" \(with no QoS)-.275 E(cons\
iderations\), "limited discrimination without QoS tiering" \(no fees ch\
ar)93.6 241 Q(ged for higher)-.198 E(-)-.22 E(quality service\), and "l\
imited discrimination and tiering" \(including higher fees allo)93.6 254
Q(wed for)-.275 E(QoS\) [NetNeutral].)93.6 267 Q
(The proponents of "netw)5.5 E(ork neutrality" are all opposed to char)
-.11 E(ging)-.198 E
(based on content \(e.g., based on applications, or the content pro)93.6
280 Q(vider\).)-.165 E(As the "netw)93.6 306 Q
(ork neutrality" discussion should mak)-.11 E 2.75(ec)-.11 G(lear)-2.75
E 2.75(,t)-.44 G(here are man)-2.75 E 2.75(yv)-.165 G(oices in the)-2.97
E(discussion that w)93.6 319 Q(ould disagree with a resource allocation\
 goal of maximizing the combined)-.11 E(aggre)93.6 332 Q -.055(ga)-.165
G(te utility).055 E 2.75(,p)-.715 G(articularly where a user')-2.75 E
2.75(su)-.605 G(tility is measured by the user')-2.75 E 2.75(sw)-.605 G
(illingness to)-2.75 E(pay)93.6 345 Q 5.5(.")-.715 G -1.21(Yo)-5.5 G
2.75(ug)1.21 G(et what you pay for" does not seem to be the consensus g\
oal for resource)-2.75 E(allocation in the Internet, either in the IETF\
 or in the commercial or political realms of the)93.6 358 Q(Internet.)
93.6 371 Q([Add citations of papers about the importance of the role of\
 the high-priced services in)93.6 397 Q
(\214nancing the infrastructure.])93.6 410 Q(Briscoe ar)93.6 436 Q
(gues for cost-f)-.198 E
(airness [B07], so that senders are made accountable for the)-.11 E
(congestion the)93.6 449 Q 2.75(yc)-.165 G 2.75(ause. There)-2.75 F
(are, of course, dif)2.75 E(ferences of opinion about ho)-.275 E 2.75
(ww)-.275 G(ell cost-)-2.75 E(based f)93.6 462 Q
(airness could be enforced, and ho)-.11 E 2.75(ww)-.275 G
(ell it \214ts the commercial reality of the Internet,)-2.75 E
(with [B07] presenting an optimistic vie)93.6 475 Q -.715(w.)-.275 G
-.44(Wi)93.6 501 S(th *only* best-ef).44 E(fort traf)-.275 E
(\214c, there w)-.275 E
(ould be fundamental limitations to the performance)-.11 E
(that real-time applications could deli)93.6 514 Q -.165(ve)-.275 G 2.75
(rt).165 G 2.75(ot)-2.75 G(heir users.)-2.75 E(In addition to the ob)5.5
E(vious needs for)-.165 E(high bandwidth, lo)93.6 527 Q 2.75(wd)-.275 G
(elay or jitter)-2.75 E 2.75(,o)-.44 G 2.75(rl)-2.75 G .55 -.275(ow p)
-2.75 H(ack).275 E(et drop rates, some applications w)-.11 E(ould lik)
-.11 E 2.75(ea)-.11 G -.11(fa)93.6 540 S(st start-up, or to be able to \
resume their old high sending rate after a relati).11 E -.165(ve)-.275 G
(ly-long idle).165 E(period, or to be able to rely on a call-setup proc\
edure so that the application is not e)93.6 553 Q -.165(ve)-.275 G(n)
.165 E(started if netw)93.6 566 Q(ork resources are not suf)-.11 E 2.75
(\214cient. There)-.275 F(are se)2.75 E -.165(ve)-.275 G
(re limitations to ho).165 E 2.75(we)-.275 G -.275(ff)-2.75 G(ecti).275
E -.165(ve)-.275 G(ly).165 E
(these requirements can be accommodated by simple best-ef)93.6 579 Q
(fort service in a congested)-.275 E(en)93.6 592 Q(vironment.)-.44 E/F1
11/Times-Bold@0 SF 2.75(2.2.2. The)72 618 R(Enf)2.75 E(or)-.275 E
(cement of F)-.198 E(air)-.275 E(ness)-.165 E F0
(As discussed in Section 3.2 belo)93.6 644 Q 1.43 -.715(w, t)-.275 H
(here are well-kno).715 E(wn problems with the enforcement of)-.275 E
-.11(fa)93.6 657 S(irness with simple best-ef).11 E(fort traf)-.275 E
(\214c \(whate)-.275 E -.165(ve)-.275 G 2.75(r").165 G -.11(fa)-2.75 G
(irness" goal is used for bandwidth).11 E
(allocation for the simple best-ef)93.6 670 Q(fort traf)-.275 E
(\214c\), and with enforcement of mechanisms for the)-.275 E -.22(avo)
93.6 683 S(idance of congestion collapse [RFC2914].).22 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
/F1 11/Times-Bold@0 SF(3.)72 85 Q/F2 14/Times-Bold@0 SF(On Flo)5.5 E
(w-Rate F)-.14 E(air)-.35 E(ness f)-.21 E(or Simple Best-Eff)-.35 E
(ort T)-.35 E(raf\214c)-1.036 E F0(This section ar)93.6 111 Q
(gues that rough \215o)-.198 E(w-rate f)-.275 E
(airness is a perfectly acceptable goal for simple)-.11 E(best-ef)93.6
124 Q(fort traf)-.275 E 2.75(\214c. This)-.275 F(section does not, ho)
2.75 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, c).165 H(laim that \215o)
.44 E(w-rate f)-.275 E(airness is necessarily)-.11 E(an *optimal* f)93.6
137 Q(airness goal or resource allocation mechanism for simple best-ef)
-.11 E(fort traf)-.275 E(\214c.)-.275 E(Simple best-ef)93.6 150 Q
(fort traf)-.275 E(\214c and \215o)-.275 E(w-rate f)-.275 E
(airness are in general not about optimality)-.11 E(.)-.715 E
(This document does *not* address an)93.6 176 Q 2.75(yi)-.165 G
(ssues about the implementation of \215o)-2.75 E(w-rate f)-.275 E
(airness.)-.11 E(In the current Internet, rough \215o)93.6 189 Q
(w-rate f)-.275 E(airness is achie)-.11 E -.165(ve)-.275 G 2.75(db).165
G 2.75(yt)-2.75 G(he f)-2.75 E(act that *most* of the)-.11 E(traf)93.6
202 Q(\214c in the Internet uses TCP)-.275 E 2.75(,a)-1.221 G
(nd *most* of the TCP connections in f)-2.75 E(act use conformant)-.11 E
(TCP congestion control [MAF05].)93.6 215 Q(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, r).165 H(ough \215o).44 E(w-rate f)-.275 E
(airness could also be achie)-.11 E -.165(ve)-.275 G(d).165 E
(by the use of per)93.6 228 Q(-\215o)-.22 E 2.75(ws)-.275 G
(cheduling at congested routers [DKS89] [LLSZ96], by related router)
-2.75 E(mechanisms [SSZ03], or by congestion-controlled transport proto\
cols other than TCP)93.6 241 Q 5.5(.N)-1.221 G(one)-5.5 E
(of those issues are addressed in this document.)93.6 254 Q
(In particular)5.5 E 2.75(,t)-.44 G(his document does not address)-2.75
E(the pros and cons of TCP-friendly congestion control, equation-based \
congestion control)93.6 267 Q([FHPW00], or an)93.6 280 Q 2.75(yo)-.165 G
2.75(ft)-2.75 G
(he myriad of other issues concerning mechanisms for approximating)-2.75
E<8d6f>93.6 293 Q(w-rate f)-.275 E 2.75(airness. Le)-.11 F(Boudec')2.75
E 2.75(st)-.605 G(utorial on rate adaption, congestion control, and f)
-2.75 E(airness)-.11 E(gi)93.6 306 Q -.165(ve)-.275 G 2.75(sa).165 G
2.75(ni)-2.75 G(ntroduction to some of these issues [B00].)-2.75 E
(In a traf)93.6 332 Q(\214c class of simple best-ef)-.275 E(fort traf)
-.275 E(\214c, it w)-.275 E(ould be possible to ha)-.11 E .33 -.165
(ve ex)-.22 H(plicit f).165 E(airness)-.11 E
(mechanisms that are implemented by the end-hosts in the netw)93.6 345 Q
(ork \(as in proportional)-.11 E -.11(fa)93.6 358 S(irness or TCP-f).11
E(airness\), e)-.11 E(xplicit f)-.165 E
(airness mechanisms enforced by the routers \(as in max-)-.11 E(min f)
93.6 371 Q(airness with F)-.11 E(air Queueing\), or a traf)-.165 E
(\214c class with no e)-.275 E(xplicit f)-.165 E(airness mechanisms at)
-.11 E(all \(as in the Internet before TCP congestion control\).)93.6
384 Q F1(3.1.)72 410 Q/F3 13/Times-Bold@0 SF(The Usefulness of Flo)5.5 E
(w-Rate F)-.13 E(air)-.325 E(ness)-.195 E F0(Man)93.6 436 Q 2.75(yo)
-.165 G 2.75(ft)-2.75 G(he useful aspects of best-ef)-2.75 E(fort traf)
-.275 E(\214c discussed abo)-.275 E .33 -.165(ve a)-.165 H
(lso quality as useful aspects).165 E(of \215o)93.6 449 Q(w-rate f)-.275
E(airness.)-.11 E(Minimal technical demands on the netw)93.6 475 Q
(ork infrastructure:)-.11 E(First, the rough \215o)115.2 491.6 Q
(w-rate f)-.275 E(airness for best-ef)-.11 E(fort traf)-.275 E
(\214c pro)-.275 E(vided by TCP or other)-.165 E
(transport protocols mak)115.2 504.6 Q
(es minimal technical demands on the infrastructure \(in the)-.11 E
(absence of policing, that is, as discussed further in the section belo)
115.2 517.6 Q 2.75(w\). The)-.275 F(rough \215o)2.75 E(w-)-.275 E
(rate f)115.2 530.6 Q(airness of TCP is pro)-.11 E
(vided by the transport protocol at the end-nodes, and doesn')-.165 E(t)
-.198 E(require an)115.2 543.6 Q 2.75(ym)-.165 G
(echanisms in routers or middlebox)-2.75 E 2.75(es. Ho)-.165 F(we)-.275
E -.165(ve)-.275 G .88 -.44(r, m).165 H(echanisms for).44 E
(enforcement of the \215o)115.2 556.6 Q(w-rate f)-.275 E(airness *w)-.11
E(ould* require some support from the)-.11 E(infrastructure.)115.2 569.6
Q(Minimal demands in terms of economic infrastructure:)93.6 586.2 Q
(Second, a system based on rough \215o)115.2 602.8 Q(w-rate f)-.275 E
(airness for best-ef)-.11 E(fort traf)-.275 E(\214c mak)-.275 E
(es minimal)-.11 E(demands in terms of economic relationships among ISP\
s or between users and ISPs.)115.2 615.8 Q(Usefulness in the real w)93.6
632.4 Q(orld:)-.11 E(Third, the current system based on rough \215o)
115.2 649 Q(w-rate f)-.275 E(airness and best-ef)-.11 E(fort traf)-.275
E(\214c has)-.275 E(sho)115.2 662 Q(wn its usefulness in the real w)
-.275 E 2.75(orld. In)-.11 F(particular)2.75 E 2.75(,t)-.44 G
(he current Internet is lar)-2.75 E(gely based)-.198 E
(of the rough \215o)115.2 675 Q(w-rate f)-.275 E(airness pro)-.11 E
(vided by TCP with best-ef)-.165 E(fort traf)-.275 E
(\214c, and the Internet)-.275 E(is still chugging along, ho)115.2 688 Q
(we)-.275 E -.165(ve)-.275 G 2.75(ri).165 G(mperfectly)-2.75 E(.)-.715 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
(Getting a share of the a)93.6 85 Q -.275(va)-.22 G(ilable bandwidth:)
.275 E(In addition, a system based on rough \215o)115.2 101.6 Q
(w-rate f)-.275 E(airness with best-ef)-.11 E(fort traf)-.275 E
(\214c gi)-.275 E -.165(ve)-.275 G 2.75(sa).165 G(ll)-2.75 E
(users a reasonable chance of getting a share of the a)115.2 114.6 Q
-.275(va)-.22 G(ilable bandwidth.).275 E(This seems to)5.5 E
(be a quality that is much appreciated by today')115.2 127.6 Q 2.75(sI)
-.605 G(nternet users.)-2.75 E/F1 11/Times-Bold@0 SF(3.2.)72 157.2 Q/F2
13/Times-Bold@0 SF(The Limitations of Flo)5.5 E(w-Rate F)-.13 E(air)
-.325 E(ness)-.195 E F0
(This section discusses some of the limitations of \215o)93.6 183.2 Q
(w-rate f)-.275 E(airness for simple best-ef)-.11 E(fort)-.275 E(traf)
93.6 196.2 Q 2.75(\214c. W)-.275 F 2.75(ew)-.88 G
(ould note that the limitations of \215o)-2.86 E(w-rate f)-.275 E
(airness are man)-.11 E 1.43 -.715(y, w)-.165 H(ith a long history).715
E(in the literature, while there is not much to be said about the bene\
\214ts of \215o)93.6 209.2 Q(w-rate f)-.275 E(airness.)-.11 E
(This does *not* mean that \215o)93.6 222.2 Q(w-rate f)-.275 E
(airness is without bene\214t.)-.11 E -.165(Fo)5.5 G 2.75(rs).165 G
(imple best-ef)-2.75 E(fort traf)-.275 E<8c63>-.275 E(with rough \215o)
93.6 235.2 Q(w-rate f)-.275 E(airness, the quote from W)-.11 E
(inston Churchill about democrac)-.44 E 2.75(yc)-.165 G(omes to)-2.75 E
(mind: "It has been said that democrac)93.6 248.2 Q 2.75(yi)-.165 G 2.75
(st)-2.75 G(he w)-2.75 E(orst form of go)-.11 E -.165(ve)-.165 G
(rnment e).165 E(xcept all the others)-.165 E(that ha)93.6 261.2 Q .33
-.165(ve b)-.22 H(een tried.").165 E(One of the ob)93.6 287.2 Q
(vious limitations of \215o)-.165 E(w-rate f)-.275 E(airness is the dif)
-.11 E(\214culty of enforcement.)-.275 E(If an)5.5 E(infrastructure is \
designed from the start with a requirement for ubiquitous per)93.6 300.2
Q(-\215o)-.22 E(w)-.275 E(scheduling, \215o)93.6 313.2 Q(w-rate f)-.275
E(airness enforced by a scheduling mechanism in routers might be)-.11 E
2.75(viable. Ho)93.6 326.2 R(we)-.275 E -.165(ve)-.275 G .88 -.44(r, w)
.165 H(hen starting with an infrastructure such as the current Internet\
 with best-).44 E(ef)93.6 339.2 Q(fort traf)-.275 E(\214c lar)-.275 E
(gely serv)-.198 E
(ed by FIFO \(First-In First-Out\) scheduling in routers and a design)
-.165 E(preference for intelligence at the ends, enforcement of \215o)
93.6 352.2 Q(w-rate f)-.275 E(airness is dif)-.11 E(\214cult at best.)
-.275 E(Further)93.6 365.2 Q 2.75(,at)-.44 G
(ransition to an infrastructure that pro)-2.75 E(vides actual \215o)
-.165 E(w-rate f)-.275 E(airness for best-ef)-.11 E(fort)-.275 E(traf)
93.6 378.2 Q(\214c enforced in routers w)-.275 E(ould be dif)-.11 E
(\214cult.)-.275 E 2.75(As)93.6 404.2 S(econd possibility)-2.75 E 2.75
(,w)-.715 G(hich is lar)-2.75 E(gely ho)-.198 E 2.75(wt)-.275 G
(he current Internet is operated, w)-2.75 E(ould be best-)-.11 E(ef)93.6
417.2 Q(fort traf)-.275 E(\214c where most of the connections, pack)
-.275 E(ets, and bytes belong to connections using)-.11 E(similar conge\
stion-control mechanisms \(in this case, those of TCP congestion contro\
l\), with)93.6 430.2 Q(fe)93.6 443.2 Q 2.75(wi)-.275 G 2.75(fa)-2.75 G
.33 -.165(ny e)-2.75 H(nforcement mechanisms.).165 E
(Of course, when this happens, the result is a rough)5.5 E
(approximation of \215o)93.6 456.2 Q(w-rate f)-.275 E
(airness, with no guarantees that the best-ef)-.11 E(fort traf)-.275 E
(\214c will)-.275 E(continue to be dominated by connections using simil\
ar congestion-control mechanisms or)93.6 469.2 Q
(that users or applications cannot g)93.6 482.2 Q
(ame the system for their bene\214t.)-.055 E(That is our current state)
5.5 E(of af)93.6 495.2 Q -.11(fa)-.275 G 2.75(irs. The).11 F(good ne)
2.75 E
(ws is that the current Internet continues to successfully carry traf)
-.275 E<8c63>-.275 E(for man)93.6 508.2 Q 2.75(yu)-.165 G 2.75(sers. In)
-2.75 F(particular)2.75 E 2.75(,w)-.44 G 2.75(ea)-2.75 G(re not a)-2.75
E -.11(wa)-.165 G(re of reports of frequent congestion collapse, or).11
E(of the Internet being dominated by se)93.6 521.2 Q -.165(ve)-.275 G
(re congestion or intolerable unf).165 E(airness.)-.11 E 2.75(At)93.6
547.2 S(hird possibility w)-2.75 E(ould be best-ef)-.11 E(fort traf)
-.275 E(\214c with \215o)-.275 E(w-rate f)-.275 E(airness pro)-.11 E
(vided by the)-.165 E
(congestion control mechanisms in the transport protocols, with some le)
93.6 560.2 Q -.165(ve)-.275 G 2.75(lo).165 G 2.75(fe)-2.75 G
(nforcement,)-2.75 E(either in congested routers, in middlebox)93.6
573.2 Q(es, or by other mechanisms.)-.165 E([Add citations to some)5.5 E
(of the literature on this.])93.6 586.2 Q
(There seems to us to be considerable promise that incenti)5.5 E -.165
(ve)-.275 G(s).165 E(among the v)93.6 599.2 Q(arious players \(ISPs, v)
-.275 E(endors, customers, standards bodies, political entities,)-.165 E
(etc.\) will align some)93.6 612.2 Q
(what, and that further progress will be made on the deplo)-.275 E
(yment of)-.11 E -.275(va)93.6 625.2 S
(rious enforcement mechanisms for \215o).275 E(w-rate f)-.275 E
(airness in best-ef)-.11 E(fort traf)-.275 E 2.75(\214c. Of)-.275 F
(course, this)2.75 E(is not lik)93.6 638.2 Q
(ely to turn in to a fully-reliable and ubiquitous enforcement of \215o)
-.11 E(w-rate f)-.275 E(airness, or)-.11 E(of an)93.6 651.2 Q 2.75(yr)
-.165 G(elated f)-2.75 E(airness goals, for best-ef)-.11 E(fort traf)
-.275 E(\214c, so this is not lik)-.275 E(ely to be satisf)-.11 E
(actory to)-.11 E(purists in this area.)93.6 664.2 Q
(The precise de\214nition of \215o)93.6 690.2 Q(w-based f)-.275 E
(airness: A second limitation of \215o)-.11 E(w-based f)-.275 E
(airness is)-.11 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
(that there is seemingly no consensus within the research, standards, o\
r technical)93.6 85 Q(communities about the precise form of \215o)93.6
98 Q(w-based f)-.275 E(airness that should be desired for best-)-.11 E
(ef)93.6 111 Q(fort traf)-.275 E 2.75(\214c. This)-.275 F(area is v)2.75
E(ery much still in \215ux, as applications, transport protocols, and t\
he)-.165 E(Internet infrastructure e)93.6 124 Q -.22(vo)-.275 G(lv).22 E
(e.)-.165 E(Some of the areas where there are range of opinions about t\
he desired goals for rough \215o)93.6 150 Q(w-)-.275 E(based f)93.6 163
Q(airness in best-ef)-.11 E(fort traf)-.275 E(\214c include the follo)
-.275 E(wing:)-.275 E 2.75(*G)93.6 189 S
(ranularity: What is the appropriate granularity)-2.75 E 2.75(,f)-.715 G
(or rough \215o)-2.75 E(w-based f)-.275 E(airness [RFC2914]?)-.11 E(Wha\
t controls should be embedded in end-to-end protocols about the number \
of connections)93.6 202 Q(opened by a single session?)93.6 215 Q 2.75
(*R)93.6 241 S(TT)-3.41 E(-f)-1.012 E
(airness: What is the desired relationship between \215o)-.11 E 2.75(wb)
-.275 G(andwidth and round-trip)-2.75 E(times, for best-ef)93.6 254 Q
(fort traf)-.275 E 2.75(\214c? As)-.275 F(sho)2.75 E
(wn in Section 3.3 of [FJ92], it w)-.275 E(ould be straightforw)-.11 E
(ard)-.11 E(to modify TCP')93.6 267 Q 2.75(sc)-.605 G
(ongestion control so that \215o)-2.75 E(ws with similar pack)-.275 E
(et drop rates b)-.11 E(ut dif)-.22 E(ferent)-.275 E(round-trip times w)
93.6 280 Q(ould recei)-.11 E .33 -.165(ve r)-.275 H
(oughly the same throughput.).165 E(It remains an open question)5.5 E
(ho)93.6 293 Q(we)-.275 E -.165(ve)-.275 G 2.75(r\().165 G
(to this author)-2.75 E 2.75(,a)-.44 G 2.75(ta)-2.75 G .33 -.165(ny r)
-2.75 H(ate\), what w).165 E(ould be the desired relationship between)
-.11 E(throughput and round-trip times for best-ef)93.6 306 Q(fort traf)
-.275 E(\214c [HSMK98].)-.275 E 2.75(*M)93.6 332 S(ultiple congested ro\
uters: What is the desired relationship between \215o)-2.75 E 2.75(wb)
-.275 G(andwidth and)-2.75 E
(the number of congested routers along the path, for best-ef)93.6 345 Q
(fort traf)-.275 E 2.75(\214c? It)-.275 F(is well established)2.75 E
(that for TCP traf)93.6 358 Q(\214c in particular)-.275 E 2.75<2c8d>-.44
G -.275(ow)-2.75 G 2.75(st).275 G(hat tra)-2.75 E -.165(ve)-.22 G
(rse multiple congested routers recei).165 E .33 -.165(ve a)-.275 H
(higher pack)93.6 371 Q(et drop rate, and therefore lo)-.11 E
(wer throughput, than \215o)-.275 E(ws with the same round-trip)-.275 E
(time that tra)93.6 384 Q -.165(ve)-.22 G
(rse only one congested router [F91].).165 E
(There is also a long-standing debate)5.5 E(between max-min f)93.6 397 Q
(airness and proportional f)-.11 E
(airness, and no consensus within the research)-.11 E
(community on the desired f)93.6 410 Q(airness goals in this area.)-.11
E 2.75(*B)93.6 436 S(ursty vs. smooth traf)-2.75 E
(\214c: What is the desired relationship between \215o)-.275 E 2.75(wb)
-.275 G(andwidth and the)-2.75 E -.22(bu)93.6 449 S
(rstiness in the sending rate of the \215o).22 E 2.75(w? Is)-.275 F
(it a goal for a b)2.75 E(ursty \215o)-.22 E 2.75(wt)-.275 G 2.75(or)
-2.75 G(ecei)-2.75 E .33 -.165(ve t)-.275 H(he same).165 E -2.475 -.22
(av e)93.6 462 T(rage bandwidth as a \215o).22 E 2.75(ww)-.275 G
(ith a smooth sending rate?)-2.75 E(The same maximum bandwidth?)5.5 E
(Ho)93.6 475 Q 2.75(wd)-.275 G
(oes the goal depend on the time scale of the b)-2.75 E
(urstiness of the b)-.22 E(ursty \215o)-.22 E 2.75(w[)-.275 G(K96]?)
-2.75 E(\(E.g., a \215o)93.6 488 Q 2.75(wt)-.275 G(hat is b)-2.75 E
(ursty on time scales of less than a round-trip time has dif)-.22 E
(ferent)-.275 E(dynamics than a \215o)93.6 501 Q 2.75(wt)-.275 G
(hat is b)-2.75 E(ursty on a time scale of seconds or minutes.\))-.22 E
(\(Section 3.2 of)5.5 E([FJ92] sho)93.6 514 Q(ws that the use of Acti)
-.275 E .33 -.165(ve Q)-.275 H(ueue Management in routers can impro).165
E .33 -.165(ve t)-.165 H(he bias).165 E(ag)93.6 527 Q(ainst b)-.055 E
(ursty traf)-.22 E(\214c found in netw)-.275 E(orks with Drop-T)-.11 E
(ail queues.\))-.88 E 2.75(*P)93.6 553 S(ack)-2.915 E
(ets or bytes: Should the rough f)-.11 E
(airness goals be in terms of pack)-.11 E(ets per second, or in)-.11 E
(bytes per second?)93.6 566 Q 2.75(*U)93.6 592 S
(nicast vs. multicast: What should the f)-2.75 E
(airness goals be between unicast and multicast)-.11 E(traf)93.6 605 Q
(\214c?)-.275 E
(There is a range of literature for each of these topics, and we ha)93.6
631 Q .33 -.165(ve n)-.22 H(ot attempted to cite it all).165 E(abo)93.6
644 Q -.165(ve)-.165 G 5.5(.R).165 G(ough \215o)-5.5 E(w-based f)-.275 E
(airness for simple best-ef)-.11 E(fort traf)-.275 E(\214c could e)-.275
E -.22(vo)-.275 G(lv).22 E 2.75(ew)-.165 G(ith a range of)-2.75 E
(possibilities for f)93.6 657 Q
(airness in terms of round-trip times, the number of congested routers,)
-.11 E(pack)93.6 670 Q(et size, or the number of recei)-.11 E -.165(ve)
-.275 G(rs per \215o).165 E 4.18 -.715(w. \()-.275 H
(Further discussion can be found in).715 E([METRICS].\))93.6 683 Q(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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
-.165(Fa)93.6 85 S(irness o).165 E -.165(ve)-.165 G 2.75(rt).165 G(ime:)
-2.75 E(One issue raised in [B07] concerns ho)93.6 111 Q 2.75(wf)-.275 G
(airness should be inte)-2.86 E(grated o)-.165 E -.165(ve)-.165 G 2.75
(rt).165 G 2.75(ime. F)-2.75 F(or)-.165 E -.165(ex)93.6 124 S
(ample, for simple best-ef).165 E(fort traf)-.275 E
(\214c, should long \215o)-.275 E(ws recei)-.275 E .33 -.165(ve l)-.275
H(ess bandwidth in bits per).165 E(second than short \215o)93.6 137 Q
5.5(ws? F)-.275 F(or cost-based f)-.165 E(airness or for QoS-based traf)
-.11 E(\214c, it seems perfectly)-.275 E(viable for there to be some sc\
enarios where the cost is a function of \215o)93.6 150 Q 2.75(wo)-.275 G
2.75(rs)-2.75 G(ession lifetime.)-2.75 E(It also seems viable for there\
 to be some scenarios where the cost of QoS-enabled traf)93.6 163 Q
(\214c is)-.275 E(independent of \215o)93.6 176 Q 2.75(wo)-.275 G 2.75
(rs)-2.75 G(ession lifetime \(e.g., for a pri)-2.75 E -.275(va)-.275 G
(te Intranet that is measured only by).275 E
(the bandwidth of the access link, b)93.6 189 Q(ut where an)-.22 E 2.75
(yt)-.165 G(raf)-2.75 E(\214c sent of that Intranet is guaranteed to)
-.275 E(recei)93.6 202 Q .33 -.165(ve a c)-.275 H(ertain QoS\).).165 E
(Ho)93.6 228 Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H
(or simple best-ef).44 E(fort traf)-.275 E
(\214c, the current form of rough f)-.275 E(airness that is not)-.11 E
(inte)93.6 241 Q(grated o)-.165 E -.165(ve)-.165 G 2.75(rt).165 G
(ime seems perfectly acceptable.)-2.75 E
(That is, in the current Internet, a user who)5.5 E
(opens a single TCP connection for ten hours *might* recei)93.6 254 Q
.33 -.165(ve t)-.275 H(he same a).165 E -.165(ve)-.22 G
(rage throughput in).165 E(bits per second, during that TCP connection,\
 as a user who opens a single TCP connection)93.6 267 Q
(for ten minutes and then goes of)93.6 280 Q 2.75(f-line. Similarly)
-.275 F 2.75(,au)-.715 G(ser who is on-line for ten hours each)-2.75 E
(day *might* recei)93.6 293 Q .33 -.165(ve t)-.275 H
(he same throughput in bits per second, and pay roughly the same cost,)
.165 E(as a user who is on-line for ten minutes each day)93.6 306 Q 5.5
(.T)-.715 G(hat seems perfectly acceptable to us.)-5.5 E
(Other pricing mechanisms between users and ISPs seem acceptable also.)
93.6 319 Q/F1 11/Times-Bold@0 SF(4.)72 345 Q/F2 14/Times-Bold@0 SF
(On the Dif\214culties of Incr)5.5 E(emental Deployment)-.252 E F0
(One of the adv)93.6 371 Q(antages of simple best-ef)-.275 E
(fort service is that it is here, today)-.275 E 2.75(,i)-.715 G 2.75(nt)
-2.75 G(he current)-2.75 E(Internet, along with the rough \215o)93.6 384
Q(w-rate f)-.275 E(airness that results from dominance of TCP')-.11 E(s)
-.605 E(congestion control.)93.6 397 Q
(While additional classes of service w)93.6 423 Q
(ould clearly be of use in the Internet, the deplo)-.11 E(yment)-.11 E
(dif)93.6 436 Q(\214culties ha)-.275 E .33 -.165(ve b)-.22 H
(een non-tri).165 E(vial [B03].)-.275 E(These dif)5.5 E
(\214culties in deplo)-.275 E(ying interlocking changes)-.11 E(to the i\
nfrastructure \(as opposed to changes in applications\) do not necessar\
ily ha)93.6 449 Q .33 -.165(ve a)-.22 H 2.75(ne).165 G(asy)-2.75 E 2.75
(\214x; the)93.6 462 R 2.75(ys)-.165 G(tem in part from the underlying \
architecture of the Internet, which has its)-2.75 E(adv)93.6 475 Q
(antages as well as its disadv)-.275 E 2.75(antages. As)-.275 F -.165
(ex)2.75 G(plained in RFC 1958 on "Architectural).165 E
(Principles of the Internet": "F)93.6 488 Q(ortunately)-.165 E 2.75(,n)
-.715 G(obody o)-2.75 E(wns the Internet, there is no centralized)-.275
E(control, and nobody can turn it of)93.6 501 Q 2.75(f[)-.275 G
(RFC1958].")-2.75 E(The dif)93.6 527 Q(\214culties of deplo)-.275 E
(yment for end-to-end intserv or dif)-.11 E(fserv mechanisms are well-)
-.275 E(kno)93.6 540 Q(wn, ha)-.275 E(ving in part to do with the dif)
-.22 E(\214culties of deplo)-.275 E(yment for the economic)-.11 E
(infrastructure that w)93.6 553 Q(ould be needed [B03].)-.11 E
(It seems lik)5.5 E(ely that cost-based pricing based on)-.11 E
(re-ECN could also ha)93.6 566 Q .33 -.165(ve a d)-.22 H(if).165 E
(\214cult deplo)-.275 E(yment path, in)-.11 E -.22(vo)-.44 G
(lving the deplo).22 E(yment of ECN-)-.11 E(marking at routers, policer\
s at both ends of a connection, and a comple)93.6 579 Q 2.75(xs)-.165 G
(et of economic)-2.75 E(relationships [B07].)93.6 592 Q -.165(Pa)93.6
618 S(pers on some of the dif).165 E(\214culties of making changes in t\
he Internet infrastructure, including)-.275 E(the dif)93.6 631 Q
(\214culties imposed by the political and economic conte)-.275 E
(xt, include [CMB07].)-.165 E(The)5.5 E(dif)93.6 644 Q(\214culty of mak\
ing changes to the Internet infrastructure is in contrast to the compar\
ati)-.275 E -.165(ve)-.275 G
(ease in making changes in Internet applications.)93.6 657 Q
([Add citations about the deplo)93.6 683 Q(yment dif)-.11 E
(\214culties of QoS, IPv6, multicast, ECN, and other)-.275 E
(mechanisms that ha)93.6 696 Q .33 -.165(ve c)-.22 H(hick).165 E
(en-and-e)-.11 E(gg deplo)-.165 E(yment problems.])-.11 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
/F1 11/Times-Bold@0 SF(5.)72 85 Q/F2 14/Times-Bold@0 SF(Related W)5.5 E
(ork)-1.05 E F1(5.1.)72 111 Q/F3 13/Times-Bold@0 SF(Fr)5.5 E
(om the IETF)-.234 E F0
(This section discusses IETF documents relating to best-ef)93.6 137 Q
(fort service and \215o)-.275 E(w-rate f)-.275 E(airness.)-.11 E
(RFC 896 on congestion control: Nagle')93.6 163 Q 2.75(sR)-.605 G
(FC 896 on "Congestion Control in IP/TCP", from)-2.75 E
(1984, raises the issue of congestion collapse, and says that "impro)
93.6 176 Q -.165(ve)-.165 G 2.75(dh).165 G(andling of congestion)-2.75 E
(is no)93.6 189 Q 2.75(wm)-.275 G(andatory" [RFC896].)-2.75 E(RFC 896 w)
5.5 E(as written in the conte)-.11 E(xt of a hea)-.165 E(vily-loaded)
-.22 E(netw)93.6 202 Q(ork, the only pri)-.11 E -.275(va)-.275 G
(te TCP/IP long-haul netw).275 E(ork in e)-.11 E
(xistence at the time \(that of F)-.165 E(ord)-.165 E(Motor Compan)93.6
215 Q 1.43 -.715(y, i)-.165 H 2.75(n1).715 G 2.75(984\). In)-2.75 F
(addition to introducing the Nagle algorithm for minimizing)2.75 E
(the transmission of small pack)93.6 228 Q(ets in TCP)-.11 E 2.75(,R)
-1.221 G(FC 896 considers the ef)-2.75 E(fecti)-.275 E -.165(ve)-.275 G
(ness of ICMP).165 E
(Source Quench for congestion control, and comments that future g)93.6
241 Q(ate)-.055 E -.11(wa)-.275 G(ys should be capable).11 E
(of defending themselv)93.6 254 Q(es ag)-.165 E
(ainst obnoxious or malicious hosts.)-.055 E(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, R).165 H(FC 896 does not).44 E
(raise the question of f)93.6 267 Q
(airness between competing users or \215o)-.11 E(ws.)-.275 E(RFC 2914 o\
n congestion control principles: RFC 2914, a Best Current Practice docu\
ment)93.6 293 Q(from 2000 on "Congestion Control Principles", discusses\
 the issues of pre)93.6 306 Q -.165(ve)-.275 G(nting congestion).165 E
(collapse, maintaining some form of f)93.6 319 Q(airness for best-ef)
-.11 E(fort traf)-.275 E(\214c, and optimizing a \215o)-.275 E(w')-.275
E(s)-.605 E(performance in terms of throughput, delay)93.6 332 Q 2.75
(,a)-.715 G(nd loss for the \215o)-2.75 E 2.75(wi)-.275 G 2.75(nq)-2.75
G 2.75(uestion. In)-2.75 F(the)2.75 E(discussion of f)93.6 345 Q
(airness, RFC 2914 outlines polic)-.11 E 2.75(yi)-.165 G
(ssues concerning the appropriate)-2.75 E(granularity of a "\215o)93.6
358 Q(w", and ackno)-.275 E
(wledges that end nodes can easily open multiple)-.275 E
(concurrent \215o)93.6 371 Q(ws to the same destination.)-.275 E
(RFC 2914 also discusses open issues concerning)5.5 E -.11(fa)93.6 384 S
(irness between reliable unicast, unreliable unicast, reliable multicas\
t and unreliable).11 E(multicast transport protocols.)93.6 397 Q
(RFC 3714 on the amorphous problem of f)93.6 423 Q
(airness: Section 3.3 of RFC 3714, an Informational)-.11 E(document fro\
m the IAB \(Internet Architecture Board\) discussing congestion control\
 for best-)93.6 436 Q(ef)93.6 449 Q(fort v)-.275 E(oice traf)-.22 E
(\214c, has a discussion of "the amorphous problem of f)-.275 E
(airness", discussing)-.11 E(complicating issues of pack)93.6 462 Q
(et sizes, round-trip times, application-le)-.11 E -.165(ve)-.275 G 2.75
(lf).165 G(unctionality)-2.75 E 2.75(,a)-.715 G(nd the)-2.75 E(lik)93.6
475 Q 2.75(e[)-.11 G(RFC3714].)-2.75 E(RFC 2616 on opening multiple con\
nections: RFC 2616, the standards track document for)93.6 501 Q(HTTP/1.\
1, speci\214es that "clients that use persistent connections SHOULD lim\
it the number)93.6 514 Q(of simultaneous connections that the)93.6 527 Q
2.75(ym)-.165 G(aintain to a gi)-2.75 E -.165(ve)-.275 G 2.75(ns).165 G
(erv)-2.75 E(er" [RFC2616] \(Section)-.165 E(8.1.4.\).)93.6 540 Q
(RFC 2309 on unresponsi)93.6 566 Q .33 -.165(ve \215)-.275 H -.275(ow)
.165 G(s: RFC 2309, an Informational document from the End-to-).275 E(E\
nd Research Group on "Recommendations on Queue Management and Congestio\
n)93.6 579 Q -1.342 -.814(Av o)93.6 592 T
(idance in the Internet" in 2000, contains the follo).814 E
(wing recommendation: "It is ur)-.275 E(gent to)-.198 E(be)93.6 605 Q
(gin or continue research, engineering, and measurement ef)-.165 E
(forts contrib)-.275 E(uting to the design)-.22 E
(of mechanisms to deal with \215o)93.6 618 Q(ws that are unresponsi)
-.275 E .33 -.165(ve t)-.275 H 2.75(oc).165 G
(ongestion noti\214cation or are)-2.75 E(responsi)93.6 631 Q .33 -.165
(ve b)-.275 H(ut more aggressi)-.055 E .33 -.165(ve t)-.275 H(han TCP)
.165 E(." [RFC2309])-1.221 E
(RFCs on QoS: There is a long history in the IETF of the de)93.6 657 Q
-.165(ve)-.275 G(lopment of QoS mechanisms).165 E(for inte)93.6 670 Q
(grated and dif)-.165 E(ferentiated services [RFC2212, RFC2475].)-.275 E
(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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
/F1 11/Times-Bold@0 SF(5.2.)72 85 Q/F2 13/Times-Bold@0 SF(Fr)5.5 E
(om Elsewher)-.234 E(e)-.234 E F0
(This section brie\215y mentions some of the man)93.6 111 Q 2.75(yp)
-.165 G(apers in the literature on best-ef)-2.75 E(fort traf)-.275 E
<8c63>-.275 E(or on f)93.6 124 Q(airness for competing \215o)-.11 E
(ws or users.)-.275 E([B07] also has a section on some of the)5.5 E
(literature re)93.6 137 Q -.055(ga)-.165 G(rding f).055 E
(airness in the Internet.)-.11 E -.165(Fa)93.6 163 S
(irness with AIMD: F).165 E(airness with AIMD \(Additi)-.165 E .33 -.165
(ve I)-.275 H(ncrease Multiplicati).165 E .33 -.165(ve D)-.275 H
(ecrease\)).165 E(congestion control w)93.6 176 Q
(as studied by Chiu and Jain in 1987, where f)-.11 E
(airness is maximized when)-.11 E(each user or \215o)93.6 189 Q 2.75(wg)
-.275 G(ets equal allocations of the bottleneck bandwidth [CJ89].)-2.75
E -1.221(Va)5.5 G 2.75(nJ)1.221 G(acobson')-2.75 E(s)-.605 E
(1988 paper on "Congestion A)93.6 202 Q -.22(vo)-.814 G
(idance and Control" de\214ned TCP').22 E 2.75(sA)-.605 G
(IMD-based congestion)-2.75 E(control mechanisms.)93.6 215 Q -.165(Fa)
93.6 241 S(ir Queueing: The 1989 paper of F).165 E
(air Queueing by Demers et al.)-.165 E(promoted F)5.5 E(air Queueing)
-.165 E(scheduling at routers as pro)93.6 254 Q(viding f)-.165 E
(air allocation of bandwidth, lo)-.11 E(wer delay for lo)-.275 E(w-)
-.275 E(bandwidth traf)93.6 267 Q(\214c, and protection from ill-beha)
-.275 E -.165(ve)-.22 G 2.75(ds).165 G(ources [DKS89].)-2.75 E(Congesti\
on-based pricing: One of the early papers on congestion-based pricing i\
n netw)93.6 293 Q(orks)-.11 E
(is the 1993 paper on "Pricing the Internet" by MacKie-Mason and V)93.6
306 Q(arian [MV93].)-1.221 E(This)5.5 E(paper proposed a "Smart Mark)
93.6 319 Q(et" to price congestion in real time, with a per)-.11 E
(-pack)-.22 E(et-char)-.11 E(ge)-.198 E(re\215ecting mar)93.6 332 Q
(ginal congestion costs.)-.198 E(Frank K)5.5 E(elly')-.275 E 2.75(sw)
-.605 G(eb page at [Proportional] has citations)-2.75 E
(to papers on proportional f)93.6 345 Q
(airness, including [K97] on "Char)-.11 E(ging and Rate Control for)
-.198 E(Elastic T)93.6 358 Q(raf)-.385 E(\214c".)-.275 E
(Other papers on pricing in computer netw)93.6 384 Q
(orks include [SCEH96], which is in part a critique)-.11 E
(of some of the pricing proposals in the literature at the time.)93.6
397 Q([SCEH96] ar)5.5 E(gues that usage)-.198 E(char)93.6 410 Q
(ges must remain at signi\214cant le)-.198 E -.165(ve)-.275 G(ls e).165
E -.165(ve)-.275 G 2.75(ni).165 G 2.75(fc)-2.75 G(ongestion is e)-2.75 E
(xtremely lo)-.165 E -.715(w.)-.275 G F1(6.)72 436 Q/F3 14/Times-Bold@0
SF(Security Considerations)5.5 E F0(This document does not propose an)
93.6 462 Q 2.75(yn)-.165 G .55 -.275(ew m)-2.75 H
(echanisms for the Internet, and so does not).275 E(require an)93.6 475
Q 2.75(ys)-.165 G(ecurity considerations.)-2.75 E F1(7.)72 501 Q F3(IAN)
5.5 E 3.5(AC)-.28 G(onsiderations)-3.5 E F0(There are no IAN)93.6 527 Q
2.75(Ac)-.385 G(onsiderations in this document.)-2.75 E F1(8.)72 553 Q
F3(Conclusions)5.5 E F1(9.)72 579 Q F3(Ackno)5.5 E(wledgements)-.14 E
(Inf)72 605 Q(ormati)-.35 E .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252
E F0 54.087([B00] J.-Y)98.7 621.6 R 2.75(.L)-1.419 G 2.75(eB)-2.75 G
(oudec, Rate adaptation, Congestion Control and F)-2.75 E(airness: A)
-.165 E -.495(Tu)181.2 634.6 S(torial, 2000.).495 E
(URL "http://citeseer)5.5 E(.ist.psu.edu/boudec00rate.html".)-.605 E
54.087([B03] G.)98.7 651.2 R(Bell, F)2.75 E(ailure to Thri)-.165 E -.165
(ve)-.275 G 2.75(:Q).165 G(oS and the Culture of Operational)-2.75 E
(Netw)181.2 664.2 Q(orking, La)-.11 E(wrence Berk)-.165 E(ele)-.11 E
2.75(yN)-.165 G(ational Laboratory)-2.75 E 2.75(,S)-.715 G
(eptember 2003.)-2.75 E 54.087([B07] B.)98.7 680.8 R(Briscoe, Flo)2.75 E
2.75(wR)-.275 G(ate F)-2.75 E
(airness: Dismantling a Religion, internet-draft)-.165 E
(draft-briscoe-tsv)181.2 693.8 Q(area-f)-.275 E(air)-.11 E(-01.txt, w)
-.22 E(ork in progress, March 2007.)-.11 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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
49.808([CJ89] Chiu,)98.7 85 R
(D.-M., and Jain, R., Analysis of the Increase and Decrease)2.75 E
(Algorithms for Congestion A)181.2 98 Q -.22(vo)-.814 G
(idance in Computer Netw).22 E(orks, Computer)-.11 E(Netw)181.2 111 Q
(orks and ISDN Systems, V)-.11 E 2.75(.1)-1.419 G(7, pp. 1-14, 1989.)
-2.75 E([The DEC)5.5 E -.77(Te)181.2 124 S(chnical Report DEC-TR-509 w)
.77 E(as in 1987.])-.11 E 36.971([CMB07] kc)98.7 140.6 R(claf)2.75 E(fy)
-.275 E 2.75(,S)-.715 G(ascha D. Meinrath, and Scott O. Bradner)-2.75 E
2.75(,T)-.44 G(he \(un\)Economic)-2.75 E
(Internet?, Internet Economics T)181.2 153.6 Q(rack, 2007.)-.385 E
39.424([DKS89] A.)98.7 170.2 R(Demers, S. K)2.75 E(esha)-.275 E 1.43
-.715(v, a)-.22 H(nd S. Shenk).715 E(er)-.11 E 2.75(,A)-.44 G
(nalysis and Simulation of a F)-2.75 E(air)-.165 E
(Queueing Algorithm, SIGCOMM, 1989.)181.2 183.2 Q 55.308([F91] Flo)98.7
199.8 R(yd, S., Connections with Multiple Congested Gate)-.11 E -.11(wa)
-.275 G(ys in P).11 E(ack)-.165 E(et-)-.11 E(Switched Netw)181.2 212.8 Q
(orks P)-.11 E(art 1: One-w)-.165 E(ay T)-.11 E(raf)-.385 E
(\214c, Computer Communication)-.275 E(Re)181.2 225.8 Q(vie)-.275 E 1.43
-.715(w, V)-.275 H(ol.21, No.5, October 1991.)-.704 E 30.866
([FHPW00] Flo)98.7 242.4 R(yd, S., Handle)-.11 E 1.43 -.715(y, M)-.165 H
(., P).715 E(adh)-.165 E(ye, J., and W)-.055 E(idmer)-.44 E 2.75(,J)-.44
G 2.75(,E)-2.75 G(quation-Based)-2.75 E
(Congestion Control for Unicast Applications, SIGCOMM, August 2000.)
181.2 255.4 Q 51.029([FJ92] On)98.7 272 R -.385(Tr)2.75 G(af).385 E
(\214c Phase Ef)-.275 E(fects in P)-.275 E(ack)-.165 E(et-Switched Gate)
-.11 E -.11(wa)-.275 G(ys, Flo).11 E(yd, S. and)-.11 E(Jacobson, V)181.2
285 Q(., Internetw)-1.419 E(orking: Research and Experience, V)-.11 E
(.3 N.3,)-1.419 E(September 1992.)181.2 298 Q 29.645
([HSMK98] Henderson,)98.7 314.6 R -.814(T.)2.75 G
(R., E. Sahouria, S. McCanne, and R.H. Katz, "On).814 E(Impro)181.2
327.6 Q(ving the F)-.165 E(airness of TCP Congestion A)-.165 E -.22(vo)
-.814 G(idance," Globecom,).22 E(No)181.2 340.6 Q -.165(ve)-.165 G
(mber 1998.).165 E 16.214([Internet2020] Internet)98.7 357.2 R(Society)
2.75 E 2.75(,")-.715 G(An Internet 2020 Initiati)-2.75 E -.165(ve)-.275
G 2.75(:").165 G(The Internet is \(still\) for)-2.75 E(Ev)181.2 370.2 Q
(eryone", 2007.)-.165 E(URL "http://www)5.5 E(.isoc.or)-.715 E(g/)-.198
E(or)181.2 383.2 Q(gs/ac/cms/uploads/docs/2020_vision.pdf".)-.198 E
53.482([K96] F)98.7 399.8 R 2.75(.K)-.88 G(elly)-3.025 E 2.75(,C)-.715 G
(har)-2.75 E(ging and Accounting for Bursty Connections, In L. W)-.198 E
(.)-1.012 E(McKnight and J. P)181.2 412.8 Q 2.75(.B)-1.221 G(aile)-2.75
E 1.43 -.715(y, e)-.165 H(ditors, Internet Economics. MIT Press,).715 E
(1997.)181.2 425.8 Q 53.482([K97] F)98.7 442.4 R 2.75(.K)-.88 G(elly)
-3.025 E 2.75(,C)-.715 G(har)-2.75 E
(ging and Rate Control for Elastic T)-.198 E(raf)-.385 E
(\214c, European)-.275 E -.385(Tr)181.2 455.4 S(ansactions on T).385 E
(elecommunications, 8:33--37, 1997.)-.77 E 35.145([LLSZ96] C.)98.7 472 R
(Lefelhocz, B. L)2.75 E(yles, S. Shenk)-.605 E(er)-.11 E 2.75(,a)-.44 G
(nd L. Zhang, Congestion Control)-2.75 E(for Best-ef)181.2 485 Q
(fort Service: Wh)-.275 E 2.75(yW)-.055 G 2.75(eN)-3.63 G(eed a Ne)-2.75
E 2.75(wP)-.275 G(aradigm, IEEE Netw)-2.915 E(ork,)-.11 E -.22(vo)181.2
498 S(l. 10, pp. 10-19, Jan. 1996.).22 E 37.587([MAF05] A.)98.7 514.6 R
(Medina, M. Allman, and S. Flo)2.75 E(yd, Measuring the Ev)-.11 E
(olution of)-.22 E -.385(Tr)181.2 527.6 S
(ansport Protocols in the Internet, Computer Communications Re).385 E
(vie)-.275 E -.715(w,)-.275 G(April 2005.)181.2 540.6 Q 24.75
([METRICS] S.)98.7 557.2 R(Flo)2.75 E(yd, Metrics for the Ev)-.11 E
(aluation of Congestion Control Mechanisms,)-.275 E
(internet-draft draft-irtf-tmr)181.2 570.2 Q(g-metrics-09.txt, w)-.198 E
(ork in progress, March)-.11 E(2007.)181.2 583.2 Q 43.703([MV93] J.)98.7
599.8 R(K. Mackie-Mason and H. V)2.75 E
(arian, "Pricing the Internet', in the)-1.221 E
(conference on Public Access to the Internet, JFK School of Go)181.2
612.8 Q -.165(ve)-.165 G(rnment,).165 E(May 1993.)181.2 625.8 Q 23.551
([NetNeutral] Netw)98.7 642.4 R(ork Neutrality)-.11 E 2.75(,W)-.715 G
2.75(ikipedia. URL)-3.19 F("http://en.wikipedia.or)181.2 655.4 Q 2.75
(g/wiki/Net_neutrality". [Added)-.198 F(for its citations to)2.75 E
(the literature.])181.2 668.4 Q(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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
17.424([Proportional] K)98.7 85 R(elly)-.275 E 2.75(,F)-.715 G
(., papers on Proportional F)-3.63 E(airness.)-.165 E 35.134
([RFC896] Nagle,)98.7 101.6 R(J., Congestion Control in IP/TCP)2.75 E
2.75(,R)-1.221 G(FC 896, January 1984.)-2.75 E 29.634([RFC1958] B.)98.7
118.2 R(Carpenter)2.75 E 2.75(,A)-.44 G
(rchitectural Principles of the Internet, RFC 1958, June)-2.75 E(1996.)
181.2 131.2 Q 29.634([RFC2212] Shenk)98.7 147.8 R(er)-.11 E 2.75(,S)-.44
G(., P)-2.75 E
(artridge, C. and R. Guerin, Speci\214cation of Guaranteed)-.165 E
(Quality of Service, RFC 2212, September 1997.)181.2 160.8 Q 29.634
([RFC2309] B.)98.7 177.4 R
(Braden at al, Recommendations on Queue Management and)2.75 E
(Congestion A)181.2 190.4 Q -.22(vo)-.814 G
(idance in the Internet, RFC 2309, April 1998.).22 E 29.634
([RFC2475] Blak)98.7 207 R(e, S., Black, D., Carlson, M., Da)-.11 E
(vies, E., W)-.22 E(ang, Z.)-.88 E(and W)5.5 E 5.5(.W)-1.012 G(eiss,)
-6.38 E(An Architecture for Dif)181.2 220 Q
(ferentiated Services, RFC 2475, December 1998.)-.275 E 29.634
([RFC2616] Fielding,)98.7 236.6 R
(R., Gettys, J., Mogul, J., Frystyk, H., Masinter)2.75 E 2.75(,L)-.44 G
(., Leach, P)-2.75 E(.)-1.221 E(and T)181.2 249.6 Q 2.75(.B)-.814 G
(erners-Lee, Hyperte)-2.75 E(xt T)-.165 E
(ransfer Protocol -- HTTP/1.1, RFC)-.385 E(2616, June 1999.)181.2 262.6
Q 29.634([RFC2914] S.)98.7 279.2 R(Flo)2.75 E
(yd, Congestion Control Principles, RFC 2914, September 2000.)-.11 E
29.634([RFC3714] S.)98.7 295.8 R(Flo)2.75 E(yd, IAB Concerns Re)-.11 E
-.055(ga)-.165 G(rding Congestion Control for V).055 E(oice T)-1.419 E
(raf)-.385 E<8c63>-.275 E(in the Internet, FRC 3714, March 2004.)181.2
308.8 Q 33.308([SCEH96] Shenk)98.7 325.4 R(er)-.11 E 2.75(,D)-.44 G 2.75
(.D)-2.75 G 2.75(.C)-2.75 G
(lark, D. Estrin, and S. Herzog, Pricing in Computer)-2.75 E(Netw)181.2
338.4 Q(orks: Reshaping the Research Agenda, A)-.11 E(CM Computer)-.44 E
(Communication Re)181.2 351.4 Q(vie)-.275 E 1.43 -.715(w, v)-.275 H
(ol. 26, April 1996.).495 E 42.471([SSZ03] I.)98.7 368 R
(Stoica, S. Shenk)2.75 E(er)-.11 E 2.75(,a)-.44 G
(nd H. Zhang, Core-Stateless F)-2.75 E(air Queueing: a)-.165 E
(Scalable Architecture to Approximate F)181.2 381 Q
(air Bandwidth Allocations in)-.165 E(High-speed Netw)181.2 394 Q
(orks, IEEE/A)-.11 E(CM T)-.44 E(ransactions on Netw)-.385 E
(orking 11\(1\):)-.11 E(33-46, 2003.)181.2 407 Q/F1 14/Times-Bold@0 SF
-.7(Au)72 433 S(thors' Addr).7 E(esses)-.252 E/F2 11/Courier@0 SF
(Sally Floyd)93.6 452 Q(ICSI Center for Internet Research)93.6 465 Q
(1947 Center Street, Suite 600)93.6 478 Q(Berkeley, CA 94704)93.6 491 Q
(USA)93.6 504 Q(EMail: floyd <at> icir <dot> org)93.6 517 Q
(URL: http:/www.icir.org/floyd/)93.6 530 Q(Mark Allman)93.6 556 Q
(ICSI Center for Internet Research)93.6 569 Q
(1947 Center Street, Suite 600)93.6 582 Q(Berkeley, CA 94704-1198)93.6
595 Q(Phone: \(440\) 235-1792)93.6 608 Q(Email: mallman at icir.org)93.6
621 Q(URL: http://www.icir.org/mallman/)93.6 634 Q F0(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 40.88(-DRAFT SIMPLE)-1.012 F
(BEST EFFOR)2.75 E 2.75(TT)-.66 G 88.444(RAFFIC June)-2.75 F(2007)2.75 E
/F1 14/Times-Bold@0 SF(Full Copyright Statement)72 85 Q F0(Cop)93.6
101.6 Q(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E(This docu\
ment is subject to the rights, licenses and restrictions contained in B\
CP 78, and)93.6 118.2 Q -.165(ex)93.6 131.2 S
(cept as set forth therein, the authors retain all their rights.).165 E
(This document and the information contained herein are pro)93.6 147.8 Q
(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)93.6 160.8 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)93.6 173.8 Q 2.75(,T)
-1.419 G(HE IETF TR)-2.75 E(UST AND THE)-.44 E(INTERNET ENGINEERING T)
93.6 186.8 Q(ASK FORCE DISCLAIM ALL W)-1.023 E(ARRANTIES, EXPRESS)-1.32
E(OR IMPLIED, INCLUDING B)93.6 199.8 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)93.6 212.8 Q
(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G(NFRINGE ANY RIGHTS OR ANY)
-2.75 E(IMPLIED W)93.6 225.8 Q(ARRANTIES OF MERCHANT)-1.32 E
(ABILITY OR FITNESS FOR A)-1.023 E -1.012(PA)93.6 238.8 S -.66(RT)1.012
G(ICULAR PURPOSE.).66 E F1(Intellectual Pr)72 264.8 Q(operty)-.252 E F0
(The IETF tak)93.6 281.4 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)93.6 294.4 Q
(technology described in this document or the e)93.6 307.4 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)93.6 320.4 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)93.6 333.4 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.)93.6 346.4 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)93.6 363
Q 2.75(ya)-.165 G(ssurances of licenses to be)-2.75 E(made a)93.6 376 Q
-.275(va)-.22 G(ilable, or the result of an attempt made to obtain a ge\
neral license or permission for).275 E(the use of such proprietary righ\
ts by implementers or users of this speci\214cation can be)93.6 389 Q
(obtained from the IETF on-line IPR repository at http://www)93.6 402 Q
(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E(The IETF in)93.6 418.6 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)93.6 431.6
Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)93.6 444.6 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
93.6 457.6 Q(g.)-.198 E(Flo)72 733 Q 361.391(yd [P)-.11 F(age 13])-.165
E 0 Cg EP
%%Trailer
end
%%EOF