
From nobody Thu Jun  3 06:57:29 2021
Return-Path: <ianswett@google.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 62DA93A12B9 for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 06:57:21 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id EaLNYCEJ-jod for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 06:57:15 -0700 (PDT)
Received: from mail-wm1-x335.google.com (mail-wm1-x335.google.com [IPv6:2a00:1450:4864:20::335]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id E38913A12AE for <webtransport@ietf.org>; Thu,  3 Jun 2021 06:57:14 -0700 (PDT)
Received: by mail-wm1-x335.google.com with SMTP id z137-20020a1c7e8f0000b02901774f2a7dc4so2857779wmc.0 for <webtransport@ietf.org>; Thu, 03 Jun 2021 06:57:14 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=+M4vrW0Nlp3mEcPhKfvxFhZLNBaOuPy5WETPYWOhTxo=; b=WpyYya2CVOn89IIhMbcYCOrn4Lx5Rbxwchv0bB+qLPFIL2FX7B0r14x7xG4bscUdmf 8whaWEqFd+i5N5V5fdExTbbP3hBqLCvYd65dioIvgKsTaVUjwSl9uJ16acv2N4Ck4o1V QAsc6RN7rYIMr6svIxy9Y5cFgD3mEsPm7ouyofV0wbnHbbLVGIsoRzXcsN1cap0XdyAw dKyL7lYbSIBz1HTeyh6/+wtS9SyIvINjO0SvLnF/hIRCbbBsdAARuVqhh/r477VRAwQB e2sGU8yLd44axNvdnYgrrKQWGj60aHlyRqWmcZycL82atYVMoqpGbIuYI7F52I/02RMf 0+/g==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=+M4vrW0Nlp3mEcPhKfvxFhZLNBaOuPy5WETPYWOhTxo=; b=mXCGRClqoym0nXrxNPykPJ/hMxdXq9efGfYJN/KXwlakkbXqwxowRY+tqd+SzKFu+3 gFWIB1735IT14U7IV4kWLSCsXDoszU9Gic/ZIN+vENiXBFEtqsBxp2dtRtB++ec2hlV+ dxZ/SztZsayAbuSOgWRqH8aE0+d2lsE0GiNWusf3qZM1RbuBxlK0TCuDCut+S65GHQwr qjGucgNrfNuFNDbqqUpwdKD8hbv5Z277AMDB1FeFQ6EzHYEQMazT5huRdFOuxzyXK1RH QrB+pt1MyQ0bTplV6EZjVczWYrWHAil8Dy+uGFmMdKd6mM4VwcS5fIqbzgo9WSlGYUqd +L9Q==
X-Gm-Message-State: AOAM533kTotxp1rU69dGemB3bGTEuhsOLVKSrzKF1423tnI/b4RazEuz Ek7TTbU29caIBwGBdyOnOEZMNXiwxIu9IGxMw/CP5w==
X-Google-Smtp-Source: ABdhPJwVYURBP4uzO6UU/MoNuj9mvYCOCYhTpgeKLWOo7spTF9otteDDh0oHynHROZ1bqIa5pCY6IjNdcP92NYfQCM4=
X-Received: by 2002:a1c:4e03:: with SMTP id g3mr10158859wmh.127.1622728632138;  Thu, 03 Jun 2021 06:57:12 -0700 (PDT)
MIME-Version: 1.0
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com>
In-Reply-To: <304A9399-D211-4463-9245-C2437E318FC3@fb.com>
From: Ian Swett <ianswett@google.com>
Date: Thu, 3 Jun 2021 09:57:00 -0400
Message-ID: <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com>
To: Alan Frindell <afrind=40fb.com@dmarc.ietf.org>
Cc: Martin Thomson <mt@lowentropy.net>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>
Content-Type: multipart/alternative; boundary="000000000000b0291705c3dcf3c1"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/tjOS7S6_fTUELFzXyfFzh_8Cvy0>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 13:57:28 -0000

--000000000000b0291705c3dcf3c1
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

I definitely agree with 1a, with the caveat that the TCP version is lower
priority than the HTTP/3 version.

I'm slightly less concerned about the stream state machine than Martin, but
I think it's definitely a real concern and I think it makes sense to have a
solution fairly well fleshed out before adopting the direction and document=
.

I'm slightly more concerned about creating a TCP fallback that doesn't work
over HTTP/1.  If we're building a fallback, it seems like these people
shouldn't be excluded, since that means developers will have to build and
maintain a 3rd fallback that's not WebTransport based.  If the dream is to
be able to develop apps for WebTransport and not have a non-WebTransport
fallback, I don't see how we accomplish that without an option over HTTP/1?

Possibly this is a forcing function to convince customers to upgrade their
middleboxes to ones which support HTTP/2 and HTTP/3, but if so I'd like to
call that out explicitly.

Or maybe the physical office with its corp middleboxes is dead and my
concern is an unfounded anachronism?

Ian

On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
40fb.com@dmarc.ietf.org> wrote:

> I think the fundamental discussion here is about whether we want to
> modify/extend the HTTP/2 state machine to handle WebTransport streams, or
> whether we should build an entirely separate stream state machine for the=
m,
> and leave the H2 machinery untouched.
>
> Martin may be right that there be dragons down the path of modifying the
> H2 state machine.  The alternative strikes me as an awful lot of
> duplication -- it's close to implementing H2 again over a single HTTP
> stream, minus header processing.  That said, I think either approach can =
be
> made to work, we just need to choose the tradeoffs we want.
>
> I am remiss in not addressing the feedback regarding unidirectional strea=
m
> resets.  I think that it got lost during the bizarre hours of the last
> IETF.  I filed an issue in the draft repo to track it (
> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I think
> the right solution is to keep the unidirectional reset functionality, and
> implement it in the H2 draft using either new frames or messages, dependi=
ng
> on the direction we take regarding the state machine.
>
> During the interim, we expressed our flexibility to adapt the design here
> as the H3 design evolves (eg: stream limits), new tools become available
> (eg: CAPSULE) and as we gather implementation experience.
>
> While I think it is important we address this design decision as soon as
> possible, I don't know that it needs to block adoption.
>
> Thanks
>
> -Alan
>
> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thomson" =
<
> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>
>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <mt@lowentropy.net>
> wrote:
>     > > I'm opposed to adoption of this document.  I don't believe that i=
t
> can be made so due to some fundamental differences in how HTTP/2 and QUIC
> stream states.
>     > >
>     > > If this problem was acknowledged, then I have every confidence
> that the people listed as authors are able to find a solution.  But
> generally adoption is about the document more than the people.  Adoption
> means agreeing that the document contains approximately the right shape f=
or
> a solution.  I can't agree with that in this case.
>     >
>     > Hi Martin,
>     >
>     > Could you elaborate on the specific differences between HTTP/2 and
> QUIC
>     > stream state machines that you are concerned about?  The only one I
> am
>     > aware of is unidirectional versus bidirectional reset, and last tim=
e
> we
>     > discussed this, the easiest solution was to make HTTP/3 behave like
>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do to
>     > address that.
>
>     That's the one I'm most concerned about, but there is also the whole
> mess around inventing semantics for unidirectional streams that needs to =
be
> worked into the state machines.
>
>     For reset, sure, you could make the HTTP/3 draft worse to deal with
> the problem.  Why would you want that?  QUIC has a much better set of
> semantics here than HTTP/2.
>
>     There is also the stream state management disconnect for
> server-initiated streams, which currently is only used for server push.
> The draft currently doesn't say anywhere near as much as it would need to
> in order to specify that part.
>
>     And managing allocation of stream limits is a little scant on detail
> in the same way.  I thought we wanted independent limits, which means
> defining new accounting systems.  (Yes, this is a criticism of the HTTP/3
> draft also, but the additional effort required to use QUIC streams is
> entirely justified in that case, because you are getting something in
> exchange.)
>
>     Why bother with all of that when you can just import QUIC frame
> semantics and then serialize them on the same stream?
>
>     Yes, this all boils down to a fairly central question: do we build th=
e
> same thing for HTTP/2 and HTTP/3?  I'm not sure that we want that.  We ne=
ed
> to do a whole bunch of stuff to use QUIC (BTW, Roberto gets to say "told
> you so" again here), but doing that same work for HTTP/2 doesn't make sen=
se
> unless the changes are minimal.  I'm suggesting that the changes aren't a=
s
> minimal as the rough sketches of protocols we have might imply.
>
>     If you conclude that there are more functions needed, as I have done,
> then you might also conclude that building those same functions into two
> different protocols is not desirable, especially when there are no benefi=
ts
> to doing so for one of those protocols.
>
>     --
>     Webtransport mailing list
>     Webtransport@ietf.org
>     https://www.ietf.org/mailman/listinfo/webtransport
>
> --
> Webtransport mailing list
> Webtransport@ietf.org
> https://www.ietf.org/mailman/listinfo/webtransport
>

--000000000000b0291705c3dcf3c1
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"></div>I definitely agree with 1a, with th=
e caveat that the=C2=A0TCP version is lower priority than the HTTP/3 versio=
n.<div><br></div><div>I&#39;m slightly less concerned about the stream stat=
e machine than Martin, but I think it&#39;s definitely a real concern and I=
 think it makes sense to have a solution fairly well fleshed out before ado=
pting the direction and document.</div><div><br></div><div>I&#39;m slightly=
 more concerned about creating=C2=A0a TCP fallback that doesn&#39;t work ov=
er HTTP/1.=C2=A0 If we&#39;re building a fallback, it seems like these peop=
le shouldn&#39;t be excluded, since that means developers will have to buil=
d and maintain a 3rd fallback that&#39;s not WebTransport based.=C2=A0 If t=
he dream is to be able to develop apps for WebTransport and not have a non-=
WebTransport fallback, I don&#39;t see how we accomplish that without an op=
tion over HTTP/1?</div><div><br></div><div>Possibly this is a forcing funct=
ion to convince customers to upgrade their middleboxes to ones which suppor=
t HTTP/2 and HTTP/3, but if so I&#39;d like to call that out explicitly.</d=
iv><div><br></div><div>Or maybe the physical office with its corp middlebox=
es is dead and my concern is an unfounded anachronism?</div><div><br></div>=
<div>Ian</div><div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D=
"gmail_attr">On Tue, May 25, 2021 at 5:36 PM Alan Frindell &lt;afrind=3D<a =
href=3D"mailto:40fb.com@dmarc.ietf.org">40fb.com@dmarc.ietf.org</a>&gt; wro=
te:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px =
0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">I think the =
fundamental discussion here is about whether we want to modify/extend the H=
TTP/2 state machine to handle WebTransport streams, or whether we should bu=
ild an entirely separate stream state machine for them, and leave the H2 ma=
chinery untouched.<br>
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br>
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>

--000000000000b0291705c3dcf3c1--


From nobody Thu Jun  3 08:38:57 2021
Return-Path: <dschinazi.ietf@gmail.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id A02E93A16AF for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 08:38:55 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_NONE=-0.0001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id JwndkQv0VpbR for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 08:38:50 -0700 (PDT)
Received: from mail-pj1-x102b.google.com (mail-pj1-x102b.google.com [IPv6:2607:f8b0:4864:20::102b]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 407163A16B2 for <webtransport@ietf.org>; Thu,  3 Jun 2021 08:38:50 -0700 (PDT)
Received: by mail-pj1-x102b.google.com with SMTP id z3-20020a17090a3983b029016bc232e40bso687608pjb.4 for <webtransport@ietf.org>; Thu, 03 Jun 2021 08:38:50 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=uZN7fK2daQvN0xy8nWAfatbDpVK1U//genmRBLROw4w=; b=h6qbre+NBpq8K3XrxzlBQgPg6VPs5eMiPtUSbPB+5htx3bCyYJDwYJ9HOjQLbiaut5 TWyl7ld0N68+M8VIG9Cw1HCGmjnuPQF5LQ4+C4ys9VPhOB/TaB5wbqkcftZC9yehW0Y5 5sQwWPfAP42c4ZVJ3bddtiudoWr3/oq6/5kBqZfEuimSFH+jKynUBReuaxwl6rnp3vnE AS8YUaY2IKPzQ6ch7iOLQeEbDxiuiMkamMH4t3428siJUdmZv9qVyKcQYfrzvHI+pGxJ ZD7aAwwqeaKTKAw3wbLTe6KsrPWyhOzfN2+tgtCV8dONYVK8go8SjYwGinQNXfbfDOg3 BpMw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=uZN7fK2daQvN0xy8nWAfatbDpVK1U//genmRBLROw4w=; b=uAWdKGRuuXkimDFSv3vqO4+Ent6VsuAYH1SGQexMBDgBITiJCPvTP3GsYXFzRh7Q/C ACX3aBoZ6KCO5hgbzzDpPCHHs4MQpGwmpnOdvEenOkGHLVqushso+ztsH5pW1M+80rww EQwfHExZxjEdM0pxtOnITQ6SEy5qMEaGNKUSCLvrioCRLBV5LqgVKP/UJJKYE9Fg1ZAM DQeLscPWXwpmsqRy8dDx5sfubuo2y7B/MmehPpveVdCo7e1lXJCTIYOciB4iCczMWAas peLuBwq0W0Hil0Ygi2qKECba+l/a6Zvv+o4Yd7iYotap2sPifol1Od9tnwTLFtgFWzVp WteA==
X-Gm-Message-State: AOAM530I1/n+PGG7ne01n0qBD2XU4ABMbMNO9JyTd55M1Qi05Q8/MKWG UgyHUiIbQM0Rt3khrIj0AG6NGzyYMDEpBMNHrcQ=
X-Google-Smtp-Source: ABdhPJxBF+goTKYQRJjtu8V4KIIKWMmNSqKsBvKXWxEI6EJjclJ4c7OqFnRgXUp8ogfEk9Xx6nwfamGr46qv1giVD+0=
X-Received: by 2002:a17:902:e9cb:b029:101:cebc:b8d with SMTP id 11-20020a170902e9cbb0290101cebc0b8dmr279460plk.5.1622734728804; Thu, 03 Jun 2021 08:38:48 -0700 (PDT)
MIME-Version: 1.0
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com> <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com>
In-Reply-To: <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com>
From: David Schinazi <dschinazi.ietf@gmail.com>
Date: Thu, 3 Jun 2021 08:38:37 -0700
Message-ID: <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com>
To: Ian Swett <ianswett@google.com>
Cc: Alan Frindell <afrind@fb.com>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>, Martin Thomson <mt@lowentropy.net>
Content-Type: multipart/alternative; boundary="000000000000133c9705c3de5fec"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/ksYd4BESwf6lsN4XtD9P8xZtJO0>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 15:38:56 -0000

--000000000000133c9705c3de5fec
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hi Ian,

Can you elaborate on what use-case would be solved by having WebTransport
support HTTP/1 ?
- We know there are networks that block QUIC and/or UDP, those will be well
served by HTTP/2.
- We know there are networks that MITM TLS then force a transparent HTTP/1
proxy, those won't allow CONNECT nor any other form of WebTransport anyway.
I'm failing to see a use-case where WebTransport over HTTP/1 would be
useful. Can you provide an example?

Thanks,
David

On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D
40google.com@dmarc.ietf.org> wrote:

> I definitely agree with 1a, with the caveat that the TCP version is lower
> priority than the HTTP/3 version.
>
> I'm slightly less concerned about the stream state machine than Martin,
> but I think it's definitely a real concern and I think it makes sense to
> have a solution fairly well fleshed out before adopting the direction and
> document.
>
> I'm slightly more concerned about creating a TCP fallback that doesn't
> work over HTTP/1.  If we're building a fallback, it seems like these peop=
le
> shouldn't be excluded, since that means developers will have to build and
> maintain a 3rd fallback that's not WebTransport based.  If the dream is t=
o
> be able to develop apps for WebTransport and not have a non-WebTransport
> fallback, I don't see how we accomplish that without an option over HTTP/=
1?
>
> Possibly this is a forcing function to convince customers to upgrade thei=
r
> middleboxes to ones which support HTTP/2 and HTTP/3, but if so I'd like t=
o
> call that out explicitly.
>
> Or maybe the physical office with its corp middleboxes is dead and my
> concern is an unfounded anachronism?
>
> Ian
>
> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
> 40fb.com@dmarc.ietf.org> wrote:
>
>> I think the fundamental discussion here is about whether we want to
>> modify/extend the HTTP/2 state machine to handle WebTransport streams, o=
r
>> whether we should build an entirely separate stream state machine for th=
em,
>> and leave the H2 machinery untouched.
>>
>> Martin may be right that there be dragons down the path of modifying the
>> H2 state machine.  The alternative strikes me as an awful lot of
>> duplication -- it's close to implementing H2 again over a single HTTP
>> stream, minus header processing.  That said, I think either approach can=
 be
>> made to work, we just need to choose the tradeoffs we want.
>>
>> I am remiss in not addressing the feedback regarding unidirectional
>> stream resets.  I think that it got lost during the bizarre hours of the
>> last IETF.  I filed an issue in the draft repo to track it (
>> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I
>> think the right solution is to keep the unidirectional reset functionali=
ty,
>> and implement it in the H2 draft using either new frames or messages,
>> depending on the direction we take regarding the state machine.
>>
>> During the interim, we expressed our flexibility to adapt the design her=
e
>> as the H3 design evolves (eg: stream limits), new tools become available
>> (eg: CAPSULE) and as we gather implementation experience.
>>
>> While I think it is important we address this design decision as soon as
>> possible, I don't know that it needs to block adoption.
>>
>> Thanks
>>
>> -Alan
>>
>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thomson"=
 <
>> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>>
>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <mt@lowentropy.net>
>> wrote:
>>     > > I'm opposed to adoption of this document.  I don't believe that
>> it can be made so due to some fundamental differences in how HTTP/2 and
>> QUIC stream states.
>>     > >
>>     > > If this problem was acknowledged, then I have every confidence
>> that the people listed as authors are able to find a solution.  But
>> generally adoption is about the document more than the people.  Adoption
>> means agreeing that the document contains approximately the right shape =
for
>> a solution.  I can't agree with that in this case.
>>     >
>>     > Hi Martin,
>>     >
>>     > Could you elaborate on the specific differences between HTTP/2 and
>> QUIC
>>     > stream state machines that you are concerned about?  The only one =
I
>> am
>>     > aware of is unidirectional versus bidirectional reset, and last
>> time we
>>     > discussed this, the easiest solution was to make HTTP/3 behave lik=
e
>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do to
>>     > address that.
>>
>>     That's the one I'm most concerned about, but there is also the whole
>> mess around inventing semantics for unidirectional streams that needs to=
 be
>> worked into the state machines.
>>
>>     For reset, sure, you could make the HTTP/3 draft worse to deal with
>> the problem.  Why would you want that?  QUIC has a much better set of
>> semantics here than HTTP/2.
>>
>>     There is also the stream state management disconnect for
>> server-initiated streams, which currently is only used for server push.
>> The draft currently doesn't say anywhere near as much as it would need t=
o
>> in order to specify that part.
>>
>>     And managing allocation of stream limits is a little scant on detail
>> in the same way.  I thought we wanted independent limits, which means
>> defining new accounting systems.  (Yes, this is a criticism of the HTTP/=
3
>> draft also, but the additional effort required to use QUIC streams is
>> entirely justified in that case, because you are getting something in
>> exchange.)
>>
>>     Why bother with all of that when you can just import QUIC frame
>> semantics and then serialize them on the same stream?
>>
>>     Yes, this all boils down to a fairly central question: do we build
>> the same thing for HTTP/2 and HTTP/3?  I'm not sure that we want that.  =
We
>> need to do a whole bunch of stuff to use QUIC (BTW, Roberto gets to say
>> "told you so" again here), but doing that same work for HTTP/2 doesn't m=
ake
>> sense unless the changes are minimal.  I'm suggesting that the changes
>> aren't as minimal as the rough sketches of protocols we have might imply=
.
>>
>>     If you conclude that there are more functions needed, as I have done=
,
>> then you might also conclude that building those same functions into two
>> different protocols is not desirable, especially when there are no benef=
its
>> to doing so for one of those protocols.
>>
>>     --
>>     Webtransport mailing list
>>     Webtransport@ietf.org
>>     https://www.ietf.org/mailman/listinfo/webtransport
>>
>> --
>> Webtransport mailing list
>> Webtransport@ietf.org
>> https://www.ietf.org/mailman/listinfo/webtransport
>>
> --
> Webtransport mailing list
> Webtransport@ietf.org
> https://www.ietf.org/mailman/listinfo/webtransport
>

--000000000000133c9705c3de5fec
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hi Ian,<div><br></div><div>Can you elaborate on what use-c=
ase would be solved by having WebTransport support HTTP/1 ?</div><div>- We =
know there are networks that block QUIC and/or UDP, those will be well serv=
ed by HTTP/2.</div><div>- We know there are networks that MITM TLS then for=
ce a transparent HTTP/1 proxy, those won&#39;t allow CONNECT nor any other =
form of WebTransport anyway.</div><div>I&#39;m failing to see a use-case wh=
ere WebTransport over HTTP/1 would be useful. Can you provide an example?</=
div><div><br></div><div>Thanks,</div><div>David</div></div><br><div class=
=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 =
at 6:57 AM Ian Swett &lt;ianswett=3D<a href=3D"mailto:40google.com@dmarc.ie=
tf.org">40google.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote cla=
ss=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid =
rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"></div>=
I definitely agree with 1a, with the caveat that the=C2=A0TCP version is lo=
wer priority than the HTTP/3 version.<div><br></div><div>I&#39;m slightly l=
ess concerned about the stream state machine than Martin, but I think it&#3=
9;s definitely a real concern and I think it makes sense to have a solution=
 fairly well fleshed out before adopting the direction and document.</div><=
div><br></div><div>I&#39;m slightly more concerned about creating=C2=A0a TC=
P fallback that doesn&#39;t work over HTTP/1.=C2=A0 If we&#39;re building a=
 fallback, it seems like these people shouldn&#39;t be excluded, since that=
 means developers will have to build and maintain a 3rd fallback that&#39;s=
 not WebTransport based.=C2=A0 If the dream is to be able to develop apps f=
or WebTransport and not have a non-WebTransport fallback, I don&#39;t see h=
ow we accomplish that without an option over HTTP/1?</div><div><br></div><d=
iv>Possibly this is a forcing function to convince customers to upgrade the=
ir middleboxes to ones which support HTTP/2 and HTTP/3, but if so I&#39;d l=
ike to call that out explicitly.</div><div><br></div><div>Or maybe the phys=
ical office with its corp middleboxes is dead and my concern is an unfounde=
d anachronism?</div><div><br></div><div>Ian</div><div><br><div class=3D"gma=
il_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Tue, May 25, 2021 at 5:3=
6 PM Alan Frindell &lt;afrind=3D<a href=3D"mailto:40fb.com@dmarc.ietf.org" =
target=3D"_blank">40fb.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquo=
te class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px =
solid rgb(204,204,204);padding-left:1ex">I think the fundamental discussion=
 here is about whether we want to modify/extend the HTTP/2 state machine to=
 handle WebTransport streams, or whether we should build an entirely separa=
te stream state machine for them, and leave the H2 machinery untouched.<br>
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br>
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div>

--000000000000133c9705c3de5fec--


From nobody Thu Jun  3 08:56:37 2021
Return-Path: <ianswett@google.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id DC1223A0772 for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 08:56:35 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id zfOfrCNh2kwH for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 08:56:30 -0700 (PDT)
Received: from mail-wr1-x430.google.com (mail-wr1-x430.google.com [IPv6:2a00:1450:4864:20::430]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id A65383A0765 for <webtransport@ietf.org>; Thu,  3 Jun 2021 08:56:29 -0700 (PDT)
Received: by mail-wr1-x430.google.com with SMTP id i94so1336769wri.4 for <webtransport@ietf.org>; Thu, 03 Jun 2021 08:56:29 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=xVrQedFkMjR8Er27wwSzcrx7vpSZwP6tIkioKWfzuvs=; b=rm3KOgMMwMdid9XG0An9r8PHwaFdJ3ObA6y2HfeAsAMToCkxdn2JtLHJjLu05XtFYx OElbBgbhBx71ZFnicY1CvPmsZh0GGpgdcUrtsuUT+EIeuKD5ZyxZI7LqGQ5YDrhVEdrY Nkk6GSdqX/TN0+jguk40CXt0nSH0q7E+Gd0b82PBTkChYxm5DGD+m99gGg/TrWyq0BYN O8NQzFvn7bVgL0bHaVsv6CDGszVB69bx7W/yKP0HjNtNS2JoquCG0MrlJVuGZrRevcWM XzqSSDAlP/lSLZMJo6Te+ik54kv2q391BpmTo+O674HsZ9MQpgE3M4l6aIBhhwL5gh5k fnhQ==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=xVrQedFkMjR8Er27wwSzcrx7vpSZwP6tIkioKWfzuvs=; b=un2zgi1asK609fYSspEQvWsM7lx4o6D9gFF/gtcmbg4xZTiQ6fJzWCd2By0/fTulGp fggxmrHywAPe0jxPxU4VhbDDwKc+geoUlgxrdqbn59XBFbFasm4fnvozTqKH25newHeW 1rSKiil3SktYEcsAAX3zEAu+85rvXy9EOEQpgdlg5FVHtNrH7d6JUR7sZI/1CxP7yEsC qdNr/LTNNfLkV2ozSe1w330dniKuPynf6NUeXXiXQGkH0qKrVTphsuVuvqiSb0efiU8k vzkpAlygTqiOvJ5Me0fpawV3dFS+vqdT6SukkVFYyWCgRH/VIayHRODKdC4Ycvs6VkoZ a7GA==
X-Gm-Message-State: AOAM532VCp6Cesg1hH+C6yW9JUXmZqTI0Gr7V7o/PCzfSiPlJXyb06JR Vl8BBb1ySwjt5FvUjlueYeeZsH0bSE5HUPLazfJ7ww==
X-Google-Smtp-Source: ABdhPJzuohgwePCWTeflIBGgPR2XQwpAxT+APb4JkRYH0ffII1Qy8YCRNvrBRieH+fmYafc2Y16jyfxSf74j2VE7Aio=
X-Received: by 2002:adf:f90c:: with SMTP id b12mr634090wrr.409.1622735786246;  Thu, 03 Jun 2021 08:56:26 -0700 (PDT)
MIME-Version: 1.0
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com> <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com> <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com>
In-Reply-To: <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com>
From: Ian Swett <ianswett@google.com>
Date: Thu, 3 Jun 2021 11:56:14 -0400
Message-ID: <CAKcm_gNnQM5i7hR3V48ETnKX=WiizK7Z27w=cb3N_qLGdoB0LA@mail.gmail.com>
To: David Schinazi <dschinazi.ietf@gmail.com>
Cc: Alan Frindell <afrind@fb.com>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>, Martin Thomson <mt@lowentropy.net>
Content-Type: multipart/alternative; boundary="0000000000001acf3605c3de9eca"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/eTlKaDjlUtupiIcStE7frokLe5g>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 15:56:36 -0000

--0000000000001acf3605c3de9eca
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <dschinazi.ietf@gmail.com>
wrote:

> Hi Ian,
>
> Can you elaborate on what use-case would be solved by having WebTransport
> support HTTP/1 ?
> - We know there are networks that block QUIC and/or UDP, those will be
> well served by HTTP/2.
> - We know there are networks that MITM TLS then force a transparent HTTP/=
1
> proxy, those won't allow CONNECT nor any other form of WebTransport anywa=
y.
> I'm failing to see a use-case where WebTransport over HTTP/1 would be
> useful. Can you provide an example?
>

I think there's evidence that those two sets of networks are very similar
today, and over time I expect the set of networks that block QUIC but don't
block HTTP/2 to be tiny(ie: <1%).

But I could be wrong.

>
> Thanks,
> David
>
> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D
> 40google.com@dmarc.ietf.org> wrote:
>
>> I definitely agree with 1a, with the caveat that the TCP version is lowe=
r
>> priority than the HTTP/3 version.
>>
>> I'm slightly less concerned about the stream state machine than Martin,
>> but I think it's definitely a real concern and I think it makes sense to
>> have a solution fairly well fleshed out before adopting the direction an=
d
>> document.
>>
>> I'm slightly more concerned about creating a TCP fallback that doesn't
>> work over HTTP/1.  If we're building a fallback, it seems like these peo=
ple
>> shouldn't be excluded, since that means developers will have to build an=
d
>> maintain a 3rd fallback that's not WebTransport based.  If the dream is =
to
>> be able to develop apps for WebTransport and not have a non-WebTransport
>> fallback, I don't see how we accomplish that without an option over HTTP=
/1?
>>
>> Possibly this is a forcing function to convince customers to upgrade
>> their middleboxes to ones which support HTTP/2 and HTTP/3, but if so I'd
>> like to call that out explicitly.
>>
>> Or maybe the physical office with its corp middleboxes is dead and my
>> concern is an unfounded anachronism?
>>
>> Ian
>>
>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
>> 40fb.com@dmarc.ietf.org> wrote:
>>
>>> I think the fundamental discussion here is about whether we want to
>>> modify/extend the HTTP/2 state machine to handle WebTransport streams, =
or
>>> whether we should build an entirely separate stream state machine for t=
hem,
>>> and leave the H2 machinery untouched.
>>>
>>> Martin may be right that there be dragons down the path of modifying th=
e
>>> H2 state machine.  The alternative strikes me as an awful lot of
>>> duplication -- it's close to implementing H2 again over a single HTTP
>>> stream, minus header processing.  That said, I think either approach ca=
n be
>>> made to work, we just need to choose the tradeoffs we want.
>>>
>>> I am remiss in not addressing the feedback regarding unidirectional
>>> stream resets.  I think that it got lost during the bizarre hours of th=
e
>>> last IETF.  I filed an issue in the draft repo to track it (
>>> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I
>>> think the right solution is to keep the unidirectional reset functional=
ity,
>>> and implement it in the H2 draft using either new frames or messages,
>>> depending on the direction we take regarding the state machine.
>>>
>>> During the interim, we expressed our flexibility to adapt the design
>>> here as the H3 design evolves (eg: stream limits), new tools become
>>> available (eg: CAPSULE) and as we gather implementation experience.
>>>
>>> While I think it is important we address this design decision as soon a=
s
>>> possible, I don't know that it needs to block adoption.
>>>
>>> Thanks
>>>
>>> -Alan
>>>
>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thomson=
" <
>>> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>>>
>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <mt@lowentropy.net=
>
>>> wrote:
>>>     > > I'm opposed to adoption of this document.  I don't believe that
>>> it can be made so due to some fundamental differences in how HTTP/2 and
>>> QUIC stream states.
>>>     > >
>>>     > > If this problem was acknowledged, then I have every confidence
>>> that the people listed as authors are able to find a solution.  But
>>> generally adoption is about the document more than the people.  Adoptio=
n
>>> means agreeing that the document contains approximately the right shape=
 for
>>> a solution.  I can't agree with that in this case.
>>>     >
>>>     > Hi Martin,
>>>     >
>>>     > Could you elaborate on the specific differences between HTTP/2 an=
d
>>> QUIC
>>>     > stream state machines that you are concerned about?  The only one
>>> I am
>>>     > aware of is unidirectional versus bidirectional reset, and last
>>> time we
>>>     > discussed this, the easiest solution was to make HTTP/3 behave
>>> like
>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do to
>>>     > address that.
>>>
>>>     That's the one I'm most concerned about, but there is also the whol=
e
>>> mess around inventing semantics for unidirectional streams that needs t=
o be
>>> worked into the state machines.
>>>
>>>     For reset, sure, you could make the HTTP/3 draft worse to deal with
>>> the problem.  Why would you want that?  QUIC has a much better set of
>>> semantics here than HTTP/2.
>>>
>>>     There is also the stream state management disconnect for
>>> server-initiated streams, which currently is only used for server push.
>>> The draft currently doesn't say anywhere near as much as it would need =
to
>>> in order to specify that part.
>>>
>>>     And managing allocation of stream limits is a little scant on detai=
l
>>> in the same way.  I thought we wanted independent limits, which means
>>> defining new accounting systems.  (Yes, this is a criticism of the HTTP=
/3
>>> draft also, but the additional effort required to use QUIC streams is
>>> entirely justified in that case, because you are getting something in
>>> exchange.)
>>>
>>>     Why bother with all of that when you can just import QUIC frame
>>> semantics and then serialize them on the same stream?
>>>
>>>     Yes, this all boils down to a fairly central question: do we build
>>> the same thing for HTTP/2 and HTTP/3?  I'm not sure that we want that. =
 We
>>> need to do a whole bunch of stuff to use QUIC (BTW, Roberto gets to say
>>> "told you so" again here), but doing that same work for HTTP/2 doesn't =
make
>>> sense unless the changes are minimal.  I'm suggesting that the changes
>>> aren't as minimal as the rough sketches of protocols we have might impl=
y.
>>>
>>>     If you conclude that there are more functions needed, as I have
>>> done, then you might also conclude that building those same functions i=
nto
>>> two different protocols is not desirable, especially when there are no
>>> benefits to doing so for one of those protocols.
>>>
>>>     --
>>>     Webtransport mailing list
>>>     Webtransport@ietf.org
>>>     https://www.ietf.org/mailman/listinfo/webtransport
>>>
>>> --
>>> Webtransport mailing list
>>> Webtransport@ietf.org
>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>
>> --
>> Webtransport mailing list
>> Webtransport@ietf.org
>> https://www.ietf.org/mailman/listinfo/webtransport
>>
>

--0000000000001acf3605c3de9eca
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quote">=
<div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 11:38 AM David=
 Schinazi &lt;<a href=3D"mailto:dschinazi.ietf@gmail.com">dschinazi.ietf@gm=
ail.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"=
margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-lef=
t:1ex"><div dir=3D"ltr">Hi Ian,<div><br></div><div>Can you elaborate on wha=
t use-case would be solved by having WebTransport support HTTP/1 ?</div><di=
v>- We know there are networks that block QUIC and/or UDP, those will be we=
ll served by HTTP/2.</div><div>- We know there are networks that MITM TLS t=
hen force a transparent HTTP/1 proxy, those won&#39;t allow CONNECT nor any=
 other form of WebTransport anyway.</div><div>I&#39;m failing to see a use-=
case where WebTransport over HTTP/1 would be useful. Can you provide an exa=
mple?</div></div></blockquote><div><br></div><div>I think there&#39;s evide=
nce that those two sets of networks are very similar today, and over time I=
 expect the set of networks that block QUIC but don&#39;t block HTTP/2 to b=
e tiny(ie: &lt;1%).</div><div><br></div><div>But I could be wrong.</div><bl=
ockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-lef=
t:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><br></=
div><div>Thanks,</div><div>David</div></div><br><div class=3D"gmail_quote">=
<div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 6:57 AM Ian Sw=
ett &lt;ianswett=3D<a href=3D"mailto:40google.com@dmarc.ietf.org" target=3D=
"_blank">40google.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote cl=
ass=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid=
 rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"></div=
>I definitely agree with 1a, with the caveat that the=C2=A0TCP version is l=
ower priority than the HTTP/3 version.<div><br></div><div>I&#39;m slightly =
less concerned about the stream state machine than Martin, but I think it&#=
39;s definitely a real concern and I think it makes sense to have a solutio=
n fairly well fleshed out before adopting the direction and document.</div>=
<div><br></div><div>I&#39;m slightly more concerned about creating=C2=A0a T=
CP fallback that doesn&#39;t work over HTTP/1.=C2=A0 If we&#39;re building =
a fallback, it seems like these people shouldn&#39;t be excluded, since tha=
t means developers will have to build and maintain a 3rd fallback that&#39;=
s not WebTransport based.=C2=A0 If the dream is to be able to develop apps =
for WebTransport and not have a non-WebTransport fallback, I don&#39;t see =
how we accomplish that without an option over HTTP/1?</div><div><br></div><=
div>Possibly this is a forcing function to convince customers to upgrade th=
eir middleboxes to ones which support HTTP/2 and HTTP/3, but if so I&#39;d =
like to call that out explicitly.</div><div><br></div><div>Or maybe the phy=
sical office with its corp middleboxes is dead and my concern is an unfound=
ed anachronism?</div><div><br></div><div>Ian</div><div><br><div class=3D"gm=
ail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Tue, May 25, 2021 at 5:=
36 PM Alan Frindell &lt;afrind=3D<a href=3D"mailto:40fb.com@dmarc.ietf.org"=
 target=3D"_blank">40fb.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockqu=
ote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px=
 solid rgb(204,204,204);padding-left:1ex">I think the fundamental discussio=
n here is about whether we want to modify/extend the HTTP/2 state machine t=
o handle WebTransport streams, or whether we should build an entirely separ=
ate stream state machine for them, and leave the H2 machinery untouched.<br=
>
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br>
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div>
</blockquote></div></div>

--0000000000001acf3605c3de9eca--


From nobody Thu Jun  3 09:28:46 2021
Return-Path: <dschinazi.ietf@gmail.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id B96E73A0CDA for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 09:28:43 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.096
X-Spam-Level: 
X-Spam-Status: No, score=-2.096 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id qfOKrA0e9bNU for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 09:28:38 -0700 (PDT)
Received: from mail-pl1-x62a.google.com (mail-pl1-x62a.google.com [IPv6:2607:f8b0:4864:20::62a]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 305723A0CD5 for <webtransport@ietf.org>; Thu,  3 Jun 2021 09:28:38 -0700 (PDT)
Received: by mail-pl1-x62a.google.com with SMTP id c13so3187858plz.0 for <webtransport@ietf.org>; Thu, 03 Jun 2021 09:28:37 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=mHcZZyTEvTa/9E/d4sEFrQEDltGyhjTeFw3yhPzNz5c=; b=ZSE7VgIZFkSYrU37VIvwBsRZIA6+bwb8KCNJVVFBDBbNYluJoQ2Tprs1Vf+4rCXo51 NPpL6szEYlzjCwLZGfCXMziIDHuE1NbU7yC1+qyWqSA1z55GfloHaG4+Ser4UcYOaqZg 8Cdol8ivDVeFzlX5hGMIAKGBJCHGpgTXPKOICwY1kC3D03ehA8QwJskYc96MvBIVHsd4 c4bLt5JQQPdzevZY6XmerOKDzXn2nJtH0TDnKls1fLCTv6YEVuHAqfITeFi4ClU0eO+E wVEarOOgQk96B/NaTdbfvoqsPOTdASvsQ16m3vHHGgwUJ05Ncx2vzn1dhJPxgPtvofXs B2nw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=mHcZZyTEvTa/9E/d4sEFrQEDltGyhjTeFw3yhPzNz5c=; b=LIAZ2lKAEjw8YmJ2dpJo+XoyPTdILURceqcTiMd/ubuxtt9EglWnmzQdU42PjkoiRl PyFae42qCqmECIselAK8DOz41emFJ/6xq59u7LLDhUZ4zHWTbMMxOie1VlGUESe8+gfY HxDTRBjrxt3z2L4ZuIoDBwA1CtgCj2TI0LnEn1OtlMvRf5qixedREDgAyd1PuvMKnw9O jYmfzqX8a9dfL8jdovnuaayRklBEKeDgVW5JRT6Ds0Fw+mqaiXTCyGUnVRW2oE24UHK9 gpa80O9nw4zSqlc/8kFNIXmNUB5DePsyBUIdikArcJ4gFlIw9KdIg7XDsmk3BxmeYwnD HmMA==
X-Gm-Message-State: AOAM531/8ynmlCpH7cm3JYfjmwjDi9FmTep2XxR0HfAHPBj1vJm7SoX1 B0UmhE7aJqrFTt63vbhzElK13V8d0xt62xBwXZk=
X-Google-Smtp-Source: ABdhPJyYf6Z0PoZzdrALFl/Mhb6+EMA1AYkv+3tRXVVkHpl8z2yi7QewdiaEoG4l7MIZJQOdSbdPyu0Ihtz+NQGQT04=
X-Received: by 2002:a17:902:a5c9:b029:f7:9f7e:aa2f with SMTP id t9-20020a170902a5c9b02900f79f7eaa2fmr315415plq.54.1622737716238; Thu, 03 Jun 2021 09:28:36 -0700 (PDT)
MIME-Version: 1.0
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com> <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com> <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com> <CAKcm_gNnQM5i7hR3V48ETnKX=WiizK7Z27w=cb3N_qLGdoB0LA@mail.gmail.com>
In-Reply-To: <CAKcm_gNnQM5i7hR3V48ETnKX=WiizK7Z27w=cb3N_qLGdoB0LA@mail.gmail.com>
From: David Schinazi <dschinazi.ietf@gmail.com>
Date: Thu, 3 Jun 2021 09:28:24 -0700
Message-ID: <CAPDSy+5V8YhXVjCw1U4KcCcC+pzXZ7dBPJoYNDojEdACfnrO2Q@mail.gmail.com>
To: Ian Swett <ianswett@google.com>
Cc: Alan Frindell <afrind@fb.com>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>, Martin Thomson <mt@lowentropy.net>
Content-Type: multipart/alternative; boundary="00000000000023db9705c3df11f7"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/_OtzswEFNrWOwslTsV0Fa-xsSUM>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 16:28:44 -0000

--00000000000023db9705c3df11f7
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Can you provide an example of a network that would benefit from this scope
increase? I've yet to see a network that MITMs TLS but allows all HTTP/1.1
through unmodified. The ones that MITM and downgrade to h1 generally do it
because they then inspect and/or modify the request and response, so
getting something with WebTransport semantics to work there at all isn't
guaranteed to be possible. Perhaps a good step forward is for us to build a
version of WebTransport over h2 and then look at deployment data and
feedback to decide if another version is needed.

David

On Thu, Jun 3, 2021 at 8:56 AM Ian Swett <ianswett@google.com> wrote:

>
>
> On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <dschinazi.ietf@gmail.com>
> wrote:
>
>> Hi Ian,
>>
>> Can you elaborate on what use-case would be solved by having WebTranspor=
t
>> support HTTP/1 ?
>> - We know there are networks that block QUIC and/or UDP, those will be
>> well served by HTTP/2.
>> - We know there are networks that MITM TLS then force a transparent
>> HTTP/1 proxy, those won't allow CONNECT nor any other form of WebTranspo=
rt
>> anyway.
>> I'm failing to see a use-case where WebTransport over HTTP/1 would be
>> useful. Can you provide an example?
>>
>
> I think there's evidence that those two sets of networks are very similar
> today, and over time I expect the set of networks that block QUIC but don=
't
> block HTTP/2 to be tiny(ie: <1%).
>
> But I could be wrong.
>
>>
>> Thanks,
>> David
>>
>> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D
>> 40google.com@dmarc.ietf.org> wrote:
>>
>>> I definitely agree with 1a, with the caveat that the TCP version is
>>> lower priority than the HTTP/3 version.
>>>
>>> I'm slightly less concerned about the stream state machine than Martin,
>>> but I think it's definitely a real concern and I think it makes sense t=
o
>>> have a solution fairly well fleshed out before adopting the direction a=
nd
>>> document.
>>>
>>> I'm slightly more concerned about creating a TCP fallback that doesn't
>>> work over HTTP/1.  If we're building a fallback, it seems like these pe=
ople
>>> shouldn't be excluded, since that means developers will have to build a=
nd
>>> maintain a 3rd fallback that's not WebTransport based.  If the dream is=
 to
>>> be able to develop apps for WebTransport and not have a non-WebTranspor=
t
>>> fallback, I don't see how we accomplish that without an option over HTT=
P/1?
>>>
>>> Possibly this is a forcing function to convince customers to upgrade
>>> their middleboxes to ones which support HTTP/2 and HTTP/3, but if so I'=
d
>>> like to call that out explicitly.
>>>
>>> Or maybe the physical office with its corp middleboxes is dead and my
>>> concern is an unfounded anachronism?
>>>
>>> Ian
>>>
>>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
>>> 40fb.com@dmarc.ietf.org> wrote:
>>>
>>>> I think the fundamental discussion here is about whether we want to
>>>> modify/extend the HTTP/2 state machine to handle WebTransport streams,=
 or
>>>> whether we should build an entirely separate stream state machine for =
them,
>>>> and leave the H2 machinery untouched.
>>>>
>>>> Martin may be right that there be dragons down the path of modifying
>>>> the H2 state machine.  The alternative strikes me as an awful lot of
>>>> duplication -- it's close to implementing H2 again over a single HTTP
>>>> stream, minus header processing.  That said, I think either approach c=
an be
>>>> made to work, we just need to choose the tradeoffs we want.
>>>>
>>>> I am remiss in not addressing the feedback regarding unidirectional
>>>> stream resets.  I think that it got lost during the bizarre hours of t=
he
>>>> last IETF.  I filed an issue in the draft repo to track it (
>>>> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I
>>>> think the right solution is to keep the unidirectional reset functiona=
lity,
>>>> and implement it in the H2 draft using either new frames or messages,
>>>> depending on the direction we take regarding the state machine.
>>>>
>>>> During the interim, we expressed our flexibility to adapt the design
>>>> here as the H3 design evolves (eg: stream limits), new tools become
>>>> available (eg: CAPSULE) and as we gather implementation experience.
>>>>
>>>> While I think it is important we address this design decision as soon
>>>> as possible, I don't know that it needs to block adoption.
>>>>
>>>> Thanks
>>>>
>>>> -Alan
>>>>
>>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thomso=
n" <
>>>> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>>>>
>>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <mt@lowentropy.ne=
t>
>>>> wrote:
>>>>     > > I'm opposed to adoption of this document.  I don't believe tha=
t
>>>> it can be made so due to some fundamental differences in how HTTP/2 an=
d
>>>> QUIC stream states.
>>>>     > >
>>>>     > > If this problem was acknowledged, then I have every confidence
>>>> that the people listed as authors are able to find a solution.  But
>>>> generally adoption is about the document more than the people.  Adopti=
on
>>>> means agreeing that the document contains approximately the right shap=
e for
>>>> a solution.  I can't agree with that in this case.
>>>>     >
>>>>     > Hi Martin,
>>>>     >
>>>>     > Could you elaborate on the specific differences between HTTP/2
>>>> and QUIC
>>>>     > stream state machines that you are concerned about?  The only on=
e
>>>> I am
>>>>     > aware of is unidirectional versus bidirectional reset, and last
>>>> time we
>>>>     > discussed this, the easiest solution was to make HTTP/3 behave
>>>> like
>>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do t=
o
>>>>     > address that.
>>>>
>>>>     That's the one I'm most concerned about, but there is also the
>>>> whole mess around inventing semantics for unidirectional streams that =
needs
>>>> to be worked into the state machines.
>>>>
>>>>     For reset, sure, you could make the HTTP/3 draft worse to deal wit=
h
>>>> the problem.  Why would you want that?  QUIC has a much better set of
>>>> semantics here than HTTP/2.
>>>>
>>>>     There is also the stream state management disconnect for
>>>> server-initiated streams, which currently is only used for server push=
.
>>>> The draft currently doesn't say anywhere near as much as it would need=
 to
>>>> in order to specify that part.
>>>>
>>>>     And managing allocation of stream limits is a little scant on
>>>> detail in the same way.  I thought we wanted independent limits, which
>>>> means defining new accounting systems.  (Yes, this is a criticism of t=
he
>>>> HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams
>>>> is entirely justified in that case, because you are getting something =
in
>>>> exchange.)
>>>>
>>>>     Why bother with all of that when you can just import QUIC frame
>>>> semantics and then serialize them on the same stream?
>>>>
>>>>     Yes, this all boils down to a fairly central question: do we build
>>>> the same thing for HTTP/2 and HTTP/3?  I'm not sure that we want that.=
  We
>>>> need to do a whole bunch of stuff to use QUIC (BTW, Roberto gets to sa=
y
>>>> "told you so" again here), but doing that same work for HTTP/2 doesn't=
 make
>>>> sense unless the changes are minimal.  I'm suggesting that the changes
>>>> aren't as minimal as the rough sketches of protocols we have might imp=
ly.
>>>>
>>>>     If you conclude that there are more functions needed, as I have
>>>> done, then you might also conclude that building those same functions =
into
>>>> two different protocols is not desirable, especially when there are no
>>>> benefits to doing so for one of those protocols.
>>>>
>>>>     --
>>>>     Webtransport mailing list
>>>>     Webtransport@ietf.org
>>>>     https://www.ietf.org/mailman/listinfo/webtransport
>>>>
>>>> --
>>>> Webtransport mailing list
>>>> Webtransport@ietf.org
>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>
>>> --
>>> Webtransport mailing list
>>> Webtransport@ietf.org
>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>
>>

--00000000000023db9705c3df11f7
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Can you provide an example of a network that would benefit=
 from this scope increase? I&#39;ve yet to see a network that MITMs TLS but=
 allows all HTTP/1.1 through unmodified. The ones that MITM and downgrade t=
o h1 generally=C2=A0do it because they then inspect and/or modify the reque=
st and response, so getting something with WebTransport semantics to work t=
here at all isn&#39;t guaranteed to be possible. Perhaps a good step forwar=
d is for us to build a version of WebTransport over h2 and then look at dep=
loyment data and feedback to decide if another version is needed.<div><br><=
/div><div>David</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" =
class=3D"gmail_attr">On Thu, Jun 3, 2021 at 8:56 AM Ian Swett &lt;<a href=
=3D"mailto:ianswett@google.com">ianswett@google.com</a>&gt; wrote:<br></div=
><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border=
-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div di=
r=3D"ltr"><br></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=
=3D"gmail_attr">On Thu, Jun 3, 2021 at 11:38 AM David Schinazi &lt;<a href=
=3D"mailto:dschinazi.ietf@gmail.com" target=3D"_blank">dschinazi.ietf@gmail=
.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"mar=
gin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1=
ex"><div dir=3D"ltr">Hi Ian,<div><br></div><div>Can you elaborate on what u=
se-case would be solved by having WebTransport support HTTP/1 ?</div><div>-=
 We know there are networks that block QUIC and/or UDP, those will be well =
served by HTTP/2.</div><div>- We know there are networks that MITM TLS then=
 force a transparent HTTP/1 proxy, those won&#39;t allow CONNECT nor any ot=
her form of WebTransport anyway.</div><div>I&#39;m failing to see a use-cas=
e where WebTransport over HTTP/1 would be useful. Can you provide an exampl=
e?</div></div></blockquote><div><br></div><div>I think there&#39;s evidence=
 that those two sets of networks are very similar today, and over time I ex=
pect the set of networks that block QUIC but don&#39;t block HTTP/2 to be t=
iny(ie: &lt;1%).</div><div><br></div><div>But I could be wrong.</div><block=
quote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1=
px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><br></div=
><div>Thanks,</div><div>David</div></div><br><div class=3D"gmail_quote"><di=
v dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 6:57 AM Ian Swett=
 &lt;ianswett=3D<a href=3D"mailto:40google.com@dmarc.ietf.org" target=3D"_b=
lank">40google.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"></div>I =
definitely agree with 1a, with the caveat that the=C2=A0TCP version is lowe=
r priority than the HTTP/3 version.<div><br></div><div>I&#39;m slightly les=
s concerned about the stream state machine than Martin, but I think it&#39;=
s definitely a real concern and I think it makes sense to have a solution f=
airly well fleshed out before adopting the direction and document.</div><di=
v><br></div><div>I&#39;m slightly more concerned about creating=C2=A0a TCP =
fallback that doesn&#39;t work over HTTP/1.=C2=A0 If we&#39;re building a f=
allback, it seems like these people shouldn&#39;t be excluded, since that m=
eans developers will have to build and maintain a 3rd fallback that&#39;s n=
ot WebTransport based.=C2=A0 If the dream is to be able to develop apps for=
 WebTransport and not have a non-WebTransport fallback, I don&#39;t see how=
 we accomplish that without an option over HTTP/1?</div><div><br></div><div=
>Possibly this is a forcing function to convince customers to upgrade their=
 middleboxes to ones which support HTTP/2 and HTTP/3, but if so I&#39;d lik=
e to call that out explicitly.</div><div><br></div><div>Or maybe the physic=
al office with its corp middleboxes is dead and my concern is an unfounded =
anachronism?</div><div><br></div><div>Ian</div><div><br><div class=3D"gmail=
_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Tue, May 25, 2021 at 5:36 =
PM Alan Frindell &lt;afrind=3D<a href=3D"mailto:40fb.com@dmarc.ietf.org" ta=
rget=3D"_blank">40fb.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote=
 class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px so=
lid rgb(204,204,204);padding-left:1ex">I think the fundamental discussion h=
ere is about whether we want to modify/extend the HTTP/2 state machine to h=
andle WebTransport streams, or whether we should build an entirely separate=
 stream state machine for them, and leave the H2 machinery untouched.<br>
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br>
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div>
</blockquote></div></div>
</blockquote></div>

--00000000000023db9705c3df11f7--


From nobody Thu Jun  3 09:40:55 2021
Return-Path: <ianswett@google.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id CFAFB3A0D78 for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 09:40:53 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.598
X-Spam-Level: 
X-Spam-Status: No, score=-17.598 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id uxuPRrZRRcMt for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 09:40:48 -0700 (PDT)
Received: from mail-wm1-x32f.google.com (mail-wm1-x32f.google.com [IPv6:2a00:1450:4864:20::32f]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 72E633A0D7E for <webtransport@ietf.org>; Thu,  3 Jun 2021 09:40:48 -0700 (PDT)
Received: by mail-wm1-x32f.google.com with SMTP id f17so3798306wmf.2 for <webtransport@ietf.org>; Thu, 03 Jun 2021 09:40:48 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=npWtAJHO9qBSaCSn++P6wNHPFFk+BcNPSDRjiY3UPV8=; b=TNXnMXDt3I47QsF8Kr4X3+Bn1FIo1/RCRVY9imPtl9eh4HeiUFzWBaKW/0TaIDVLcy zzKV40lKLJe56ohLvIdb75Ho+u1YSpGHtytSmmJOi+uIklWS4XLjYXnlMmRVTbgicT+K whyjuCTNQoQve7FT5KqHQezIu+Xma0Zd8YKJY3CNBLL0ZOqnnNrBQyI54Qso5GLk/zQa GS2b7cTByMWLh/cmCQHzxze+IqYnQD29HOxP9XmdA/el2+dKC6Ooqij7nbYPUi2uTDRE Nbb6zGUYyC24ow2ARtsI2xXcmoDSg6bbHLViT5BhYZxU7gqZDb6l+ptuvHMIcjXFedmK PdUw==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=npWtAJHO9qBSaCSn++P6wNHPFFk+BcNPSDRjiY3UPV8=; b=MDZQTGJJ5hSYA5mIT0txaWEDpLUFVx6Zvr9BwRvhTYoP8e0RGya3a+4Hx6x0/3V0RZ Q0QEdjPKVre85HuzB5VFHhmvnmM7Cea4qOIuwGAGPZ/FlAbZ67JnHVgGCVDi7CwsgyyA /VUvLnVpktFb2IYgELK827GX7CDXSVmOxBZPox7UWEfVAVKjQq+KdkxvrmC7+nkqsLMm PBWLJ338wdOR6VAKJXz+ZQyhh93e/pykT3iSdR+mLm4jtXF7j1DQWybQYO6yySUOUU55 r+vAIaJzoR76Uq81QoF3mUHwx1fx6IffDzS4KCBIk+mNIUtt3p0eR29YP3dGYEtJ/coj AfKQ==
X-Gm-Message-State: AOAM5316Vnk0fGeVbGitcMrUzJV7hApXwJTocP8f34q/szpOeObJcet0 EHArUUhbXxHtiVc6qnsA4wXZZXn7/IpkdM/UHOMQcg==
X-Google-Smtp-Source: ABdhPJxBvtDPKs6o8iU7W+BDk4Dth2yB3TwMx64GTgQXWTm6so0l3mo98SE6l7L9BhYE3HePeHYTWF5IaN8CK3/zEVM=
X-Received: by 2002:a05:600c:2054:: with SMTP id p20mr11282908wmg.165.1622738445654;  Thu, 03 Jun 2021 09:40:45 -0700 (PDT)
MIME-Version: 1.0
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com> <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com> <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com> <CAKcm_gNnQM5i7hR3V48ETnKX=WiizK7Z27w=cb3N_qLGdoB0LA@mail.gmail.com> <CAPDSy+5V8YhXVjCw1U4KcCcC+pzXZ7dBPJoYNDojEdACfnrO2Q@mail.gmail.com>
In-Reply-To: <CAPDSy+5V8YhXVjCw1U4KcCcC+pzXZ7dBPJoYNDojEdACfnrO2Q@mail.gmail.com>
From: Ian Swett <ianswett@google.com>
Date: Thu, 3 Jun 2021 12:40:32 -0400
Message-ID: <CAKcm_gMF3zpGP92tAv7MeKdFR+=Vc6uvGJ2PGFjTgxFWv9eJ7w@mail.gmail.com>
To: David Schinazi <dschinazi.ietf@gmail.com>
Cc: Alan Frindell <afrind@fb.com>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>, Martin Thomson <mt@lowentropy.net>
Content-Type: multipart/alternative; boundary="0000000000009e350c05c3df3c5b"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/ONmIa1xcUOvgF5Jc3Fk7gOmSxig>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 16:40:54 -0000

--0000000000009e350c05c3df3c5b
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Thanks David,

I was thinking of corporate networks with MITMs that inspect HTTP 1.1 in a
limited way, but it sounds like those may not be very common, so there's no
way to make WebTransport work anyway.

If so, making WebTransport HTTP/2 specific is fine with me.

I still think apps that want to work close to 100% of the time won't be
able to depend upon WebTransport always working, but maybe that's just the
state of the internet.

Ian

On Thu, Jun 3, 2021 at 12:28 PM David Schinazi <dschinazi.ietf@gmail.com>
wrote:

> Can you provide an example of a network that would benefit from this scop=
e
> increase? I've yet to see a network that MITMs TLS but allows all HTTP/1.=
1
> through unmodified. The ones that MITM and downgrade to h1 generally do i=
t
> because they then inspect and/or modify the request and response, so
> getting something with WebTransport semantics to work there at all isn't
> guaranteed to be possible. Perhaps a good step forward is for us to build=
 a
> version of WebTransport over h2 and then look at deployment data and
> feedback to decide if another version is needed.
>
> David
>
> On Thu, Jun 3, 2021 at 8:56 AM Ian Swett <ianswett@google.com> wrote:
>
>>
>>
>> On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <dschinazi.ietf@gmail.com=
>
>> wrote:
>>
>>> Hi Ian,
>>>
>>> Can you elaborate on what use-case would be solved by having
>>> WebTransport support HTTP/1 ?
>>> - We know there are networks that block QUIC and/or UDP, those will be
>>> well served by HTTP/2.
>>> - We know there are networks that MITM TLS then force a transparent
>>> HTTP/1 proxy, those won't allow CONNECT nor any other form of WebTransp=
ort
>>> anyway.
>>> I'm failing to see a use-case where WebTransport over HTTP/1 would be
>>> useful. Can you provide an example?
>>>
>>
>> I think there's evidence that those two sets of networks are very simila=
r
>> today, and over time I expect the set of networks that block QUIC but do=
n't
>> block HTTP/2 to be tiny(ie: <1%).
>>
>> But I could be wrong.
>>
>>>
>>> Thanks,
>>> David
>>>
>>> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D
>>> 40google.com@dmarc.ietf.org> wrote:
>>>
>>>> I definitely agree with 1a, with the caveat that the TCP version is
>>>> lower priority than the HTTP/3 version.
>>>>
>>>> I'm slightly less concerned about the stream state machine than Martin=
,
>>>> but I think it's definitely a real concern and I think it makes sense =
to
>>>> have a solution fairly well fleshed out before adopting the direction =
and
>>>> document.
>>>>
>>>> I'm slightly more concerned about creating a TCP fallback that doesn't
>>>> work over HTTP/1.  If we're building a fallback, it seems like these p=
eople
>>>> shouldn't be excluded, since that means developers will have to build =
and
>>>> maintain a 3rd fallback that's not WebTransport based.  If the dream i=
s to
>>>> be able to develop apps for WebTransport and not have a non-WebTranspo=
rt
>>>> fallback, I don't see how we accomplish that without an option over HT=
TP/1?
>>>>
>>>> Possibly this is a forcing function to convince customers to upgrade
>>>> their middleboxes to ones which support HTTP/2 and HTTP/3, but if so I=
'd
>>>> like to call that out explicitly.
>>>>
>>>> Or maybe the physical office with its corp middleboxes is dead and my
>>>> concern is an unfounded anachronism?
>>>>
>>>> Ian
>>>>
>>>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
>>>> 40fb.com@dmarc.ietf.org> wrote:
>>>>
>>>>> I think the fundamental discussion here is about whether we want to
>>>>> modify/extend the HTTP/2 state machine to handle WebTransport streams=
, or
>>>>> whether we should build an entirely separate stream state machine for=
 them,
>>>>> and leave the H2 machinery untouched.
>>>>>
>>>>> Martin may be right that there be dragons down the path of modifying
>>>>> the H2 state machine.  The alternative strikes me as an awful lot of
>>>>> duplication -- it's close to implementing H2 again over a single HTTP
>>>>> stream, minus header processing.  That said, I think either approach =
can be
>>>>> made to work, we just need to choose the tradeoffs we want.
>>>>>
>>>>> I am remiss in not addressing the feedback regarding unidirectional
>>>>> stream resets.  I think that it got lost during the bizarre hours of =
the
>>>>> last IETF.  I filed an issue in the draft repo to track it (
>>>>> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I
>>>>> think the right solution is to keep the unidirectional reset function=
ality,
>>>>> and implement it in the H2 draft using either new frames or messages,
>>>>> depending on the direction we take regarding the state machine.
>>>>>
>>>>> During the interim, we expressed our flexibility to adapt the design
>>>>> here as the H3 design evolves (eg: stream limits), new tools become
>>>>> available (eg: CAPSULE) and as we gather implementation experience.
>>>>>
>>>>> While I think it is important we address this design decision as soon
>>>>> as possible, I don't know that it needs to block adoption.
>>>>>
>>>>> Thanks
>>>>>
>>>>> -Alan
>>>>>
>>>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thoms=
on" <
>>>>> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>>>>>
>>>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <
>>>>> mt@lowentropy.net> wrote:
>>>>>     > > I'm opposed to adoption of this document.  I don't believe
>>>>> that it can be made so due to some fundamental differences in how HTT=
P/2
>>>>> and QUIC stream states.
>>>>>     > >
>>>>>     > > If this problem was acknowledged, then I have every confidenc=
e
>>>>> that the people listed as authors are able to find a solution.  But
>>>>> generally adoption is about the document more than the people.  Adopt=
ion
>>>>> means agreeing that the document contains approximately the right sha=
pe for
>>>>> a solution.  I can't agree with that in this case.
>>>>>     >
>>>>>     > Hi Martin,
>>>>>     >
>>>>>     > Could you elaborate on the specific differences between HTTP/2
>>>>> and QUIC
>>>>>     > stream state machines that you are concerned about?  The only
>>>>> one I am
>>>>>     > aware of is unidirectional versus bidirectional reset, and last
>>>>> time we
>>>>>     > discussed this, the easiest solution was to make HTTP/3 behave
>>>>> like
>>>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do
>>>>> to
>>>>>     > address that.
>>>>>
>>>>>     That's the one I'm most concerned about, but there is also the
>>>>> whole mess around inventing semantics for unidirectional streams that=
 needs
>>>>> to be worked into the state machines.
>>>>>
>>>>>     For reset, sure, you could make the HTTP/3 draft worse to deal
>>>>> with the problem.  Why would you want that?  QUIC has a much better s=
et of
>>>>> semantics here than HTTP/2.
>>>>>
>>>>>     There is also the stream state management disconnect for
>>>>> server-initiated streams, which currently is only used for server pus=
h.
>>>>> The draft currently doesn't say anywhere near as much as it would nee=
d to
>>>>> in order to specify that part.
>>>>>
>>>>>     And managing allocation of stream limits is a little scant on
>>>>> detail in the same way.  I thought we wanted independent limits, whic=
h
>>>>> means defining new accounting systems.  (Yes, this is a criticism of =
the
>>>>> HTTP/3 draft also, but the additional effort required to use QUIC str=
eams
>>>>> is entirely justified in that case, because you are getting something=
 in
>>>>> exchange.)
>>>>>
>>>>>     Why bother with all of that when you can just import QUIC frame
>>>>> semantics and then serialize them on the same stream?
>>>>>
>>>>>     Yes, this all boils down to a fairly central question: do we buil=
d
>>>>> the same thing for HTTP/2 and HTTP/3?  I'm not sure that we want that=
.  We
>>>>> need to do a whole bunch of stuff to use QUIC (BTW, Roberto gets to s=
ay
>>>>> "told you so" again here), but doing that same work for HTTP/2 doesn'=
t make
>>>>> sense unless the changes are minimal.  I'm suggesting that the change=
s
>>>>> aren't as minimal as the rough sketches of protocols we have might im=
ply.
>>>>>
>>>>>     If you conclude that there are more functions needed, as I have
>>>>> done, then you might also conclude that building those same functions=
 into
>>>>> two different protocols is not desirable, especially when there are n=
o
>>>>> benefits to doing so for one of those protocols.
>>>>>
>>>>>     --
>>>>>     Webtransport mailing list
>>>>>     Webtransport@ietf.org
>>>>>     https://www.ietf.org/mailman/listinfo/webtransport
>>>>>
>>>>> --
>>>>> Webtransport mailing list
>>>>> Webtransport@ietf.org
>>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>>
>>>> --
>>>> Webtransport mailing list
>>>> Webtransport@ietf.org
>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>
>>>

--0000000000009e350c05c3df3c5b
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Thanks David,<div><br></div><div>I was thinking of corpora=
te networks with MITMs that inspect HTTP 1.1 in a limited way, but it sound=
s like those may not be very common, so there&#39;s no way to make WebTrans=
port work anyway.</div><div><br></div><div>If so, making WebTransport HTTP/=
2 specific is fine with me.</div><div><br></div><div>I still think apps tha=
t want to work close to 100% of the time won&#39;t be able to depend upon W=
ebTransport always working, but maybe that&#39;s just the state of the inte=
rnet.</div><div><br></div><div>Ian</div></div><br><div class=3D"gmail_quote=
"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 12:28 PM Dav=
id Schinazi &lt;<a href=3D"mailto:dschinazi.ietf@gmail.com">dschinazi.ietf@=
gmail.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=
=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding=
-left:1ex"><div dir=3D"ltr">Can you provide an example of a network that wo=
uld benefit from this scope increase? I&#39;ve yet to see a network that MI=
TMs TLS but allows all HTTP/1.1 through unmodified. The ones that MITM and =
downgrade to h1 generally=C2=A0do it because they then inspect and/or modif=
y the request and response, so getting something with WebTransport semantic=
s to work there at all isn&#39;t guaranteed to be possible. Perhaps a good =
step forward is for us to build a version of WebTransport over h2 and then =
look at deployment data and feedback to decide if another version is needed=
.<div><br></div><div>David</div></div><br><div class=3D"gmail_quote"><div d=
ir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 8:56 AM Ian Swett &l=
t;<a href=3D"mailto:ianswett@google.com" target=3D"_blank">ianswett@google.=
com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"marg=
in:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1e=
x"><div dir=3D"ltr"><div dir=3D"ltr"><br></div><br><div class=3D"gmail_quot=
e"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 11:38 AM Da=
vid Schinazi &lt;<a href=3D"mailto:dschinazi.ietf@gmail.com" target=3D"_bla=
nk">dschinazi.ietf@gmail.com</a>&gt; wrote:<br></div><blockquote class=3D"g=
mail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204=
,204,204);padding-left:1ex"><div dir=3D"ltr">Hi Ian,<div><br></div><div>Can=
 you elaborate on what use-case would be solved by having WebTransport supp=
ort HTTP/1 ?</div><div>- We know there are networks that block QUIC and/or =
UDP, those will be well served by HTTP/2.</div><div>- We know there are net=
works that MITM TLS then force a transparent HTTP/1 proxy, those won&#39;t =
allow CONNECT nor any other form of WebTransport anyway.</div><div>I&#39;m =
failing to see a use-case where WebTransport over HTTP/1 would be useful. C=
an you provide an example?</div></div></blockquote><div><br></div><div>I th=
ink there&#39;s evidence that those two sets of networks are very similar t=
oday, and over time I expect the set of networks that block QUIC but don&#3=
9;t block HTTP/2 to be tiny(ie: &lt;1%).</div><div><br></div><div>But I cou=
ld be wrong.</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px=
 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div di=
r=3D"ltr"><div><br></div><div>Thanks,</div><div>David</div></div><br><div c=
lass=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2=
021 at 6:57 AM Ian Swett &lt;ianswett=3D<a href=3D"mailto:40google.com@dmar=
c.ietf.org" target=3D"_blank">40google.com@dmarc.ietf.org</a>&gt; wrote:<br=
></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;=
border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><=
div dir=3D"ltr"></div>I definitely agree with 1a, with the caveat that the=
=C2=A0TCP version is lower priority than the HTTP/3 version.<div><br></div>=
<div>I&#39;m slightly less concerned about the stream state machine than Ma=
rtin, but I think it&#39;s definitely a real concern and I think it makes s=
ense to have a solution fairly well fleshed out before adopting the directi=
on and document.</div><div><br></div><div>I&#39;m slightly more concerned a=
bout creating=C2=A0a TCP fallback that doesn&#39;t work over HTTP/1.=C2=A0 =
If we&#39;re building a fallback, it seems like these people shouldn&#39;t =
be excluded, since that means developers will have to build and maintain a =
3rd fallback that&#39;s not WebTransport based.=C2=A0 If the dream is to be=
 able to develop apps for WebTransport and not have a non-WebTransport fall=
back, I don&#39;t see how we accomplish that without an option over HTTP/1?=
</div><div><br></div><div>Possibly this is a forcing function to convince c=
ustomers to upgrade their middleboxes to ones which support HTTP/2 and HTTP=
/3, but if so I&#39;d like to call that out explicitly.</div><div><br></div=
><div>Or maybe the physical office with its corp middleboxes is dead and my=
 concern is an unfounded anachronism?</div><div><br></div><div>Ian</div><di=
v><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On T=
ue, May 25, 2021 at 5:36 PM Alan Frindell &lt;afrind=3D<a href=3D"mailto:40=
fb.com@dmarc.ietf.org" target=3D"_blank">40fb.com@dmarc.ietf.org</a>&gt; wr=
ote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px=
 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">I think the=
 fundamental discussion here is about whether we want to modify/extend the =
HTTP/2 state machine to handle WebTransport streams, or whether we should b=
uild an entirely separate stream state machine for them, and leave the H2 m=
achinery untouched.<br>
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br>
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div>
</blockquote></div></div>
</blockquote></div>
</blockquote></div>

--0000000000009e350c05c3df3c5b--


From nobody Thu Jun  3 11:17:31 2021
Return-Path: <ekinnear@apple.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 757273A1282 for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 11:17:29 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.793
X-Spam-Level: 
X-Spam-Status: No, score=-2.793 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.698, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, HTML_MESSAGE=0.001, MIME_QP_LONG_LINE=0.001, RCVD_IN_DNSWL_BLOCKED=0.001, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=unavailable autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=apple.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id kYR2gy8MfI0g for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 11:17:25 -0700 (PDT)
Received: from ma1-aaemail-dr-lapp01.apple.com (ma1-aaemail-dr-lapp01.apple.com [17.171.2.60]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id E89D93A127D for <webtransport@ietf.org>; Thu,  3 Jun 2021 11:17:24 -0700 (PDT)
Received: from pps.filterd (ma1-aaemail-dr-lapp01.apple.com [127.0.0.1]) by ma1-aaemail-dr-lapp01.apple.com (8.16.0.42/8.16.0.42) with SMTP id 153IDrvA015940; Thu, 3 Jun 2021 11:17:19 -0700
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=apple.com; h=content-type : content-transfer-encoding : from : mime-version : subject : message-id : date : cc : to; s=20180706; bh=InfCTHRT3Qx+OmkvaG29IfGmA5HeoakcdGOEh4KgBjI=; b=gHVWZsA0/uTpIye2Ci1ILmcQ/tZ+6eeVGX3Q46WAoOfRcZIH3qJjc4vCSqLs4hIbHnI6 HZBxiEWj2tAAFwXcZjF5FHN+XV4/PDouCe4C/K1E6Af4Us1Yx7mLdSxmrfONp3Nr4cdv PtzJyhB5Rv2lIfKEzWm2fFoUtVfwamn8FiViYjknLpjtknYQCeHYqbZ7GduBm0hA7Xo0 +KyKvkmY9jMS6QC+q+jMhfFEi2Ng+YzlhZzSXcGJRfzshWCWB+/EUon4FpXGp3j0hpOo j0xezqwJ38ovJUYE0OwMncIfPLLQ4rUjSU7W9W4VIaffFehVwwKvPJsik8KoDEy6K/gM OQ== 
Received: from ma-mailsvcp-mta-lapp04.corp.apple.com (ma-mailsvcp-mta-lapp04.corp.apple.com [10.226.18.136]) by ma1-aaemail-dr-lapp01.apple.com with ESMTP id 38um947h58-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NO); Thu, 03 Jun 2021 11:17:19 -0700
Received: from ma-mailsvcp-mmp-lapp01.apple.com (ma-mailsvcp-mmp-lapp01.apple.com [17.32.222.14]) by ma-mailsvcp-mta-lapp04.corp.apple.com (Oracle Communications Messaging Server 8.1.0.9.20210415 64bit (built Apr 15 2021)) with ESMTPS id <0QU5009O71GUPL00@ma-mailsvcp-mta-lapp04.corp.apple.com>; Thu, 03 Jun 2021 11:17:18 -0700 (PDT)
Received: from process_milters-daemon.ma-mailsvcp-mmp-lapp01.apple.com by ma-mailsvcp-mmp-lapp01.apple.com (Oracle Communications Messaging Server 8.1.0.9.20210415 64bit (built Apr 15 2021)) id <0QU500S00118LL00@ma-mailsvcp-mmp-lapp01.apple.com>; Thu, 03 Jun 2021 11:17:18 -0700 (PDT)
X-Va-A: 
X-Va-T-CD: 8154b23b3511e33f7d3c4e1c60118686
X-Va-E-CD: 3a9f8742fe417345fd085d88c18259b7
X-Va-R-CD: 34a0388f460ef2b7e3117bdb230280d5
X-Va-CD: 0
X-Va-ID: 6b5a4b10-1b7d-4b68-84e8-8cc5612f7d58
X-V-A: 
X-V-T-CD: 8154b23b3511e33f7d3c4e1c60118686
X-V-E-CD: 3a9f8742fe417345fd085d88c18259b7
X-V-R-CD: 34a0388f460ef2b7e3117bdb230280d5
X-V-CD: 0
X-V-ID: 87a4238b-50bb-40e8-b508-a5b65b231b58
X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:6.0.391, 18.0.761 definitions=2021-06-03_12:2021-06-03, 2021-06-03 signatures=0
Received: from smtpclient.apple (unknown [17.10.131.213]) by ma-mailsvcp-mmp-lapp01.apple.com (Oracle Communications Messaging Server 8.1.0.9.20210415 64bit (built Apr 15 2021)) with ESMTPSA id <0QU50081A1GTXW00@ma-mailsvcp-mmp-lapp01.apple.com>; Thu, 03 Jun 2021 11:17:18 -0700 (PDT)
Content-type: multipart/alternative; boundary=Apple-Mail-6A880F41-721F-4355-9C6D-A5E5573766A6
Content-transfer-encoding: 7bit
From: Eric Kinnear <ekinnear@apple.com>
MIME-version: 1.0 (1.0)
Message-id: <7D4F22BF-6910-4E45-BC27-F395461B603C@apple.com>
Date: Thu, 03 Jun 2021 14:17:16 -0400
Cc: David Schinazi <dschinazi.ietf@gmail.com>, Victor Vasiliev <vasilvv@google.com>, WebTransport <webtransport@ietf.org>, Alan Frindell <afrind@fb.com>, Martin Thomson <mt@lowentropy.net>
To: Ian Swett <ianswett=40google.com@dmarc.ietf.org>
X-Mailer: iPad Mail (19A266a)
X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:6.0.391, 18.0.761 definitions=2021-06-03_12:2021-06-03, 2021-06-03 signatures=0
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/bFXYYmIsnWMeFFO0TfkdJoR9mlc>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 18:17:30 -0000

--Apple-Mail-6A880F41-721F-4355-9C6D-A5E5573766A6
Content-Type: text/plain;
	charset=utf-8
Content-Transfer-Encoding: quoted-printable

=EF=BB=BFI think the question of HTTP/1.1 is an interesting one: we=E2=80=99=
ve so far been operating with the view that falling back from QUIC to TLS/TC=
P might enable significantly more folks to participate, but that the delta b=
etween HTTP/1.1 and HTTP/2 would not be as significant.

I don=E2=80=99t have massively strong opinions in terms of how we go about a=
ddressing that, beyond the earlier concerns about additional effort required=
 to cover HTTP/1.1 vs. expected gain, but if the expected gain goes up, then=
 perhaps that changes things sufficiently.=20

It seems that some of that thinking has come from the observations made abou=
t _who_ is in control of the versioning; if your network operator does not a=
llow QUIC then that=E2=80=99s less under your control than whether HTTP/2 is=
 supported (although, as Ian mentioned, there=E2=80=99s a population with mi=
ddleboxes that force you down to HTTP/1.1, which might or might not support C=
ONNECT anyways).=20

Would it be possible to gather some data about the relative size of these di=
fferent populations? I=E2=80=99m not sure there=E2=80=99s anyone in a great p=
osition to do that, at least for CONNECT specifically if it=E2=80=99s not al=
ready possible/in-use, but perhaps data about other H1 vs. H2 vs. H3 usage w=
ould be interesting.

We know that, at least on iOS, HTTP/2 is in use in a higher percentage for w=
eb browsing use cases than for other application use cases.

> On Jun 3, 2021, at 12:41 PM, Ian Swett <ianswett=3D40google.com@dmarc.ietf=
.org> wrote:
> =EF=BB=BF
> Thanks David,
>=20
> I was thinking of corporate networks with MITMs that inspect HTTP 1.1 in a=
 limited way, but it sounds like those may not be very common, so there's no=
 way to make WebTransport work anyway.
>=20
> If so, making WebTransport HTTP/2 specific is fine with me.
>=20
> I still think apps that want to work close to 100% of the time won't be ab=
le to depend upon WebTransport always working, but maybe that's just the sta=
te of the internet.
>=20
> Ian
>=20
> On Thu, Jun 3, 2021 at 12:28 PM David Schinazi <dschinazi.ietf@gmail.com> w=
rote:
>> Can you provide an example of a network that would benefit from this scop=
e increase? I've yet to see a network that MITMs TLS but allows all HTTP/1.1=
 through unmodified. The ones that MITM and downgrade to h1 generally do it b=
ecause they then inspect and/or modify the request and response, so getting s=
omething with WebTransport semantics to work there at all isn't guaranteed t=
o be possible. Perhaps a good step forward is for us to build a version of W=
ebTransport over h2 and then look at deployment data and feedback to decide i=
f another version is needed.
>>=20
>> David
>>=20
>> On Thu, Jun 3, 2021 at 8:56 AM Ian Swett <ianswett@google.com> wrote:
>>>=20
>>>=20
>>> On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <dschinazi.ietf@gmail.com=
> wrote:
>>>> Hi Ian,
>>>>=20
>>>> Can you elaborate on what use-case would be solved by having WebTranspo=
rt support HTTP/1 ?
>>>> - We know there are networks that block QUIC and/or UDP, those will be w=
ell served by HTTP/2.
>>>> - We know there are networks that MITM TLS then force a transparent HTT=
P/1 proxy, those won't allow CONNECT nor any other form of WebTransport anyw=
ay.
>>>> I'm failing to see a use-case where WebTransport over HTTP/1 would be u=
seful. Can you provide an example?
>>>=20
>>> I think there's evidence that those two sets of networks are very simila=
r today, and over time I expect the set of networks that block QUIC but don'=
t block HTTP/2 to be tiny(ie: <1%).
>>>=20
>>> But I could be wrong.
>>>>=20
>>>> Thanks,
>>>> David
>>>>=20
>>>> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D40google.com@dmarc=
.ietf.org> wrote:
>>>>> I definitely agree with 1a, with the caveat that the TCP version is lo=
wer priority than the HTTP/3 version.
>>>>>=20
>>>>> I'm slightly less concerned about the stream state machine than Martin=
, but I think it's definitely a real concern and I think it makes sense to h=
ave a solution fairly well fleshed out before adopting the direction and doc=
ument.
>>>>>=20
>>>>> I'm slightly more concerned about creating a TCP fallback that doesn't=
 work over HTTP/1.  If we're building a fallback, it seems like these people=
 shouldn't be excluded, since that means developers will have to build and m=
aintain a 3rd fallback that's not WebTransport based.  If the dream is to be=
 able to develop apps for WebTransport and not have a non-WebTransport fallb=
ack, I don't see how we accomplish that without an option over HTTP/1?
>>>>>=20
>>>>> Possibly this is a forcing function to convince customers to upgrade t=
heir middleboxes to ones which support HTTP/2 and HTTP/3, but if so I'd like=
 to call that out explicitly.
>>>>>=20
>>>>> Or maybe the physical office with its corp middleboxes is dead and my c=
oncern is an unfounded anachronism?
>>>>>=20
>>>>> Ian
>>>>>=20
>>>>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D40fb.com@dmarc=
.ietf.org> wrote:
>>>>>> I think the fundamental discussion here is about whether we want to m=
odify/extend the HTTP/2 state machine to handle WebTransport streams, or whe=
ther we should build an entirely separate stream state machine for them, and=
 leave the H2 machinery untouched.

This sounds about right to me, if we think that the mapping of H3 onto H2 is=
 not how we=E2=80=99d want it to be, we=E2=80=99ve evolved it significantly s=
o far, and I don=E2=80=99t see any reason why we can=E2=80=99t continue to d=
o so.
>>>>>>=20
>>>>>> Martin may be right that there be dragons down the path of modifying t=
he H2 state machine.  The alternative strikes me as an awful lot of duplicat=
ion -- it's close to implementing H2 again over a single HTTP stream, minus h=
eader processing.  That said, I think either approach can be made to work, w=
e just need to choose the tradeoffs we want. =20
>>>>>>=20
>>>>>> I am remiss in not addressing the feedback regarding unidirectional s=
tream resets.  I think that it got lost during the bizarre hours of the last=
 IETF.  I filed an issue in the draft repo to track it (https://github.com/e=
kinnear/draft-webtransport-http2/issues/17).  I think the right solution is t=
o keep the unidirectional reset functionality, and implement it in the H2 dr=
aft using either new frames or messages, depending on the direction we take r=
egarding the state machine.
>>>>>>=20
>>>>>> During the interim, we expressed our flexibility to adapt the design h=
ere as the H3 design evolves (eg: stream limits), new tools become available=
 (eg: CAPSULE) and as we gather implementation experience.
>>>>>>=20
>>>>>> While I think it is important we address this design decision as soon=
 as possible, I don't know that it needs to block adoption.=20

I=E2=80=99d agree, I support adoption (as an author), and with that I=E2=80=99=
d very much like to sit down and make sure we=E2=80=99re covering Martin=E2=80=
=99s feedback here =E2=80=94 I know parts of it were also raised at 110 prio=
r to the interim and we haven=E2=80=99t all synced up to make sure it=E2=80=99=
s sorted out. That might result in minor changes to the document, it might r=
esult in major changes to the document =E2=80=94 either is fine from my pers=
pective.

Thanks,
Eric


>>>>>>=20
>>>>>> Thanks
>>>>>>=20
>>>>>> -Alan
>>>>>>=20
>>>>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thoms=
on" <webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>>>>>>=20
>>>>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>>>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <mt@lowentropy.n=
et> wrote:
>>>>>>     > > I'm opposed to adoption of this document.  I don't believe th=
at it can be made so due to some fundamental differences in how HTTP/2 and Q=
UIC stream states.
>>>>>>     > >=20
>>>>>>     > > If this problem was acknowledged, then I have every confidenc=
e that the people listed as authors are able to find a solution.  But genera=
lly adoption is about the document more than the people.  Adoption means agr=
eeing that the document contains approximately the right shape for a solutio=
n.  I can't agree with that in this case.
>>>>>>     >=20
>>>>>>     > Hi Martin,
>>>>>>     >=20
>>>>>>     > Could you elaborate on the specific differences between HTTP/2 a=
nd QUIC=20
>>>>>>     > stream state machines that you are concerned about?  The only o=
ne I am=20
>>>>>>     > aware of is unidirectional versus bidirectional reset, and last=
 time we=20
>>>>>>     > discussed this, the easiest solution was to make HTTP/3 behave l=
ike=20
>>>>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do t=
o=20
>>>>>>     > address that.
>>>>>>=20
>>>>>>     That's the one I'm most concerned about, but there is also the wh=
ole mess around inventing semantics for unidirectional streams that needs to=
 be worked into the state machines.
>>>>>>=20
>>>>>>     For reset, sure, you could make the HTTP/3 draft worse to deal wi=
th the problem.  Why would you want that?  QUIC has a much better set of sem=
antics here than HTTP/2.
>>>>>>=20
>>>>>>     There is also the stream state management disconnect for server-i=
nitiated streams, which currently is only used for server push.  The draft c=
urrently doesn't say anywhere near as much as it would need to in order to s=
pecify that part.
>>>>>>=20
>>>>>>     And managing allocation of stream limits is a little scant on det=
ail in the same way.  I thought we wanted independent limits, which means de=
fining new accounting systems.  (Yes, this is a criticism of the HTTP/3 draf=
t also, but the additional effort required to use QUIC streams is entirely j=
ustified in that case, because you are getting something in exchange.)
>>>>>>=20
>>>>>>     Why bother with all of that when you can just import QUIC frame s=
emantics and then serialize them on the same stream?
>>>>>>=20
>>>>>>     Yes, this all boils down to a fairly central question: do we buil=
d the same thing for HTTP/2 and HTTP/3?  I'm not sure that we want that.  We=
 need to do a whole bunch of stuff to use QUIC (BTW, Roberto gets to say "to=
ld you so" again here), but doing that same work for HTTP/2 doesn't make sen=
se unless the changes are minimal.  I'm suggesting that the changes aren't a=
s minimal as the rough sketches of protocols we have might imply.
>>>>>>=20
>>>>>>     If you conclude that there are more functions needed, as I have d=
one, then you might also conclude that building those same functions into tw=
o different protocols is not desirable, especially when there are no benefit=
s to doing so for one of those protocols.
>>>>>>=20
>>>>>>     --=20
>>>>>>     Webtransport mailing list
>>>>>>     Webtransport@ietf.org
>>>>>>     https://www.ietf.org/mailman/listinfo/webtransport=20
>>>>>>=20
>>>>>> --=20
>>>>>> Webtransport mailing list
>>>>>> Webtransport@ietf.org
>>>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>> --=20
>>>>> Webtransport mailing list
>>>>> Webtransport@ietf.org
>>>>> https://www.ietf.org/mailman/listinfo/webtransport
> --=20
> Webtransport mailing list
> Webtransport@ietf.org
> https://www.ietf.org/mailman/listinfo/webtransport

--Apple-Mail-6A880F41-721F-4355-9C6D-A5E5573766A6
Content-Type: text/html;
	charset=utf-8
Content-Transfer-Encoding: quoted-printable

<html><head><meta http-equiv=3D"content-type" content=3D"text/html; charset=3D=
utf-8"></head><body dir=3D"auto"><div dir=3D"ltr">=EF=BB=BFI think the quest=
ion of HTTP/1.1 is an interesting one: we=E2=80=99ve so far been operating w=
ith the view that falling back from QUIC to TLS/TCP might enable significant=
ly more folks to participate, but that the delta between HTTP/1.1 and HTTP/2=
 would not be as significant.<div dir=3D"ltr"><div dir=3D"ltr"><br></div><di=
v dir=3D"ltr">I don=E2=80=99t have massively strong opinions in terms of how=
 we go about addressing that, beyond the earlier concerns about additional e=
ffort required to cover HTTP/1.1 vs. expected gain, but if the expected gain=
 goes up, then perhaps that changes things sufficiently.&nbsp;</div><div dir=
=3D"ltr"><br></div><div dir=3D"ltr">It seems that some of that thinking has c=
ome from the observations made about _who_ is in control of the versioning; i=
f your network operator does not allow QUIC then that=E2=80=99s less under y=
our control than whether HTTP/2 is supported (although, as Ian mentioned, th=
ere=E2=80=99s a population with middleboxes that force you down to HTTP/1.1,=
 which might or might not support CONNECT anyways).&nbsp;</div><div dir=3D"l=
tr"><br></div><div dir=3D"ltr">Would it be possible to gather some data abou=
t the relative size of these different populations? I=E2=80=99m not sure the=
re=E2=80=99s anyone in a great position to do that, at least for CONNECT spe=
cifically if it=E2=80=99s not already possible/in-use, but perhaps data abou=
t other H1 vs. H2 vs. H3 usage would be interesting.</div><div dir=3D"ltr"><=
br></div><div dir=3D"ltr">We know that, at least on iOS, HTTP/2 is in use in=
 a higher percentage for web browsing use cases than for other application u=
se cases.</div><div dir=3D"ltr"><br><blockquote type=3D"cite">On Jun 3, 2021=
, at 12:41 PM, Ian Swett &lt;ianswett=3D40google.com@dmarc.ietf.org&gt; wrot=
e:<br><br></blockquote></div><blockquote type=3D"cite"><div dir=3D"ltr">=EF=BB=
=BF<div dir=3D"ltr">Thanks David,<div><br></div><div>I was thinking of corpo=
rate networks with MITMs that inspect HTTP 1.1 in a limited way, but it soun=
ds like those may not be very common, so there's no way to make WebTransport=
 work anyway.</div><div><br></div><div>If so, making WebTransport HTTP/2 spe=
cific is fine with me.</div><div><br></div><div>I still think apps that want=
 to work close to 100% of the time won't be able to depend upon WebTransport=
 always working, but maybe that's just the state of the internet.</div><div>=
<br></div><div>Ian</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr=
" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 12:28 PM David Schinazi &lt;<a=
 href=3D"mailto:dschinazi.ietf@gmail.com">dschinazi.ietf@gmail.com</a>&gt; w=
rote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px=
 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"=
ltr">Can you provide an example of a network that would benefit from this sc=
ope increase? I've yet to see a network that MITMs TLS but allows all HTTP/1=
.1 through unmodified. The ones that MITM and downgrade to h1 generally&nbsp=
;do it because they then inspect and/or modify the request and response, so g=
etting something with WebTransport semantics to work there at all isn't guar=
anteed to be possible. Perhaps a good step forward is for us to build a vers=
ion of WebTransport over h2 and then look at deployment data and feedback to=
 decide if another version is needed.<div><br></div><div>David</div></div><b=
r><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, J=
un 3, 2021 at 8:56 AM Ian Swett &lt;<a href=3D"mailto:ianswett@google.com" t=
arget=3D"_blank">ianswett@google.com</a>&gt; wrote:<br></div><blockquote cla=
ss=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"><br></di=
v><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Th=
u, Jun 3, 2021 at 11:38 AM David Schinazi &lt;<a href=3D"mailto:dschinazi.ie=
tf@gmail.com" target=3D"_blank">dschinazi.ietf@gmail.com</a>&gt; wrote:<br><=
/div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bor=
der-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">Hi Ia=
n,<div><br></div><div>Can you elaborate on what use-case would be solved by h=
aving WebTransport support HTTP/1 ?</div><div>- We know there are networks t=
hat block QUIC and/or UDP, those will be well served by HTTP/2.</div><div>- W=
e know there are networks that MITM TLS then force a transparent HTTP/1 prox=
y, those won't allow CONNECT nor any other form of WebTransport anyway.</div=
><div>I'm failing to see a use-case where WebTransport over HTTP/1 would be u=
seful. Can you provide an example?</div></div></blockquote><div><br></div><d=
iv>I think there's evidence that those two sets of networks are very similar=
 today, and over time I expect the set of networks that block QUIC but don't=
 block HTTP/2 to be tiny(ie: &lt;1%).</div><div><br></div><div>But I could b=
e wrong.</div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0=
.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"lt=
r"><div><br></div><div>Thanks,</div><div>David</div></div><br><div class=3D"=
gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 6:=
57 AM Ian Swett &lt;ianswett=3D<a href=3D"mailto:40google.com@dmarc.ietf.org=
" target=3D"_blank">40google.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blo=
ckquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:=
1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"lt=
r"></div>I definitely agree with 1a, with the caveat that the&nbsp;TCP versi=
on is lower priority than the HTTP/3 version.<div><br></div><div>I'm slightl=
y less concerned about the stream state machine than Martin, but I think it'=
s definitely a real concern and I think it makes sense to have a solution fa=
irly well fleshed out before adopting the direction and document.</div><div>=
<br></div><div>I'm slightly more concerned about creating&nbsp;a TCP fallbac=
k that doesn't work over HTTP/1.&nbsp; If we're building a fallback, it seem=
s like these people shouldn't be excluded, since that means developers will h=
ave to build and maintain a 3rd fallback that's not WebTransport based.&nbsp=
; If the dream is to be able to develop apps for WebTransport and not have a=
 non-WebTransport fallback, I don't see how we accomplish that without an op=
tion over HTTP/1?</div><div><br></div><div>Possibly this is a forcing functi=
on to convince customers to upgrade their middleboxes to ones which support H=
TTP/2 and HTTP/3, but if so I'd like to call that out explicitly.</div><div>=
<br></div><div>Or maybe the physical office with its corp middleboxes is dea=
d and my concern is an unfounded anachronism?</div><div><br></div><div>Ian</=
div><div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr=
">On Tue, May 25, 2021 at 5:36 PM Alan Frindell &lt;afrind=3D<a href=3D"mail=
to:40fb.com@dmarc.ietf.org" target=3D"_blank">40fb.com@dmarc.ietf.org</a>&gt=
; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0=
px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">I think th=
e fundamental discussion here is about whether we want to modify/extend the H=
TTP/2 state machine to handle WebTransport streams, or whether we should bui=
ld an entirely separate stream state machine for them, and leave the H2 mach=
inery untouched.<br></blockquote></div></div></div></blockquote></div></bloc=
kquote></div></div></blockquote></div></blockquote></div></div></blockquote>=
<div dir=3D"ltr"><br></div>This sounds about right to me, if we think that t=
he mapping of H3 onto H2 is not how we=E2=80=99d want it to be, we=E2=80=99v=
e evolved it significantly so far, and I don=E2=80=99t see any reason why we=
 can=E2=80=99t continue to do so.<br><blockquote type=3D"cite"><div dir=3D"l=
tr"><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"ma=
rgin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1=
ex"><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"ma=
rgin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1=
ex"><div dir=3D"ltr"><div class=3D"gmail_quote"><blockquote class=3D"gmail_q=
uote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,20=
4);padding-left:1ex"><div class=3D"gmail_quote"><blockquote class=3D"gmail_q=
uote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,20=
4);padding-left:1ex"><div dir=3D"ltr"><div><div class=3D"gmail_quote"><block=
quote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1p=
x solid rgb(204,204,204);padding-left:1ex">
<br>
Martin may be right that there be dragons down the path of modifying the H2 s=
tate machine.&nbsp; The alternative strikes me as an awful lot of duplicatio=
n -- it's close to implementing H2 again over a single HTTP stream, minus he=
ader processing.&nbsp; That said, I think either approach can be made to wor=
k, we just need to choose the tradeoffs we want.&nbsp; <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream r=
esets.&nbsp; I think that it got lost during the bizarre hours of the last I=
ETF.&nbsp; I filed an issue in the draft repo to track it (<a href=3D"https:=
//github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"noreferrer"=
 target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2/issu=
es/17</a>).&nbsp; I think the right solution is to keep the unidirectional r=
eset functionality, and implement it in the H2 draft using either new frames=
 or messages, depending on the direction we take regarding the state machine=
.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here as=
 the H3 design evolves (eg: stream limits), new tools become available (eg: C=
APSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as pos=
sible, I don't know that it needs to block adoption. <br></blockquote></div>=
</div></div></blockquote></div></blockquote></div></div></blockquote></div><=
/blockquote></div></div></blockquote><div dir=3D"ltr"><br></div>I=E2=80=99d a=
gree, I support adoption (as an author), and with that I=E2=80=99d very much=
 like to sit down and make sure we=E2=80=99re covering Martin=E2=80=99s feed=
back here =E2=80=94 I know parts of it were also raised at 110 prior to the i=
nterim and we haven=E2=80=99t all synced up to make sure it=E2=80=99s sorted=
 out. That might result in minor changes to the document, it might result in=
 major changes to the document =E2=80=94 either is fine from my perspective.=
</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">Thanks,</div><div dir=3D"l=
tr">Eric</div><div dir=3D"ltr"><br></div><div dir=3D"ltr"><br><blockquote ty=
pe=3D"cite"><div dir=3D"ltr"><div class=3D"gmail_quote"><blockquote class=3D=
"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(20=
4,204,204);padding-left:1ex"><div class=3D"gmail_quote"><blockquote class=3D=
"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(20=
4,204,204);padding-left:1ex"><div dir=3D"ltr"><div class=3D"gmail_quote"><bl=
ockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left=
:1px solid rgb(204,204,204);padding-left:1ex"><div class=3D"gmail_quote"><bl=
ockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left=
:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><div cla=
ss=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px=
 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thomson" &lt=
;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_blank">webtrans=
port-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@lowentropy.net" t=
arget=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
&nbsp; &nbsp; On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
&nbsp; &nbsp; &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hre=
f=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; w=
rote:<br>
&nbsp; &nbsp; &gt; &gt; I'm opposed to adoption of this document.&nbsp; I do=
n't believe that it can be made so due to some fundamental differences in ho=
w HTTP/2 and QUIC stream states.<br>
&nbsp; &nbsp; &gt; &gt; <br>
&nbsp; &nbsp; &gt; &gt; If this problem was acknowledged, then I have every c=
onfidence that the people listed as authors are able to find a solution.&nbs=
p; But generally adoption is about the document more than the people.&nbsp; A=
doption means agreeing that the document contains approximately the right sh=
ape for a solution.&nbsp; I can't agree with that in this case.<br>
&nbsp; &nbsp; &gt; <br>
&nbsp; &nbsp; &gt; Hi Martin,<br>
&nbsp; &nbsp; &gt; <br>
&nbsp; &nbsp; &gt; Could you elaborate on the specific differences between H=
TTP/2 and QUIC <br>
&nbsp; &nbsp; &gt; stream state machines that you are concerned about?&nbsp;=
 The only one I am <br>
&nbsp; &nbsp; &gt; aware of is unidirectional versus bidirectional reset, an=
d last time we <br>
&nbsp; &nbsp; &gt; discussed this, the easiest solution was to make HTTP/3 b=
ehave like <br>
&nbsp; &nbsp; &gt; HTTP/2, which would mean there's nothing HTTP/2 draft wou=
ld do to <br>
&nbsp; &nbsp; &gt; address that.<br>
<br>
&nbsp; &nbsp; That's the one I'm most concerned about, but there is also the=
 whole mess around inventing semantics for unidirectional streams that needs=
 to be worked into the state machines.<br>
<br>
&nbsp; &nbsp; For reset, sure, you could make the HTTP/3 draft worse to deal=
 with the problem.&nbsp; Why would you want that?&nbsp; QUIC has a much bett=
er set of semantics here than HTTP/2.<br>
<br>
&nbsp; &nbsp; There is also the stream state management disconnect for serve=
r-initiated streams, which currently is only used for server push.&nbsp; The=
 draft currently doesn't say anywhere near as much as it would need to in or=
der to specify that part.<br>
<br>
&nbsp; &nbsp; And managing allocation of stream limits is a little scant on d=
etail in the same way.&nbsp; I thought we wanted independent limits, which m=
eans defining new accounting systems.&nbsp; (Yes, this is a criticism of the=
 HTTP/3 draft also, but the additional effort required to use QUIC streams i=
s entirely justified in that case, because you are getting something in exch=
ange.)<br>
<br>
&nbsp; &nbsp; Why bother with all of that when you can just import QUIC fram=
e semantics and then serialize them on the same stream?<br>
<br>
&nbsp; &nbsp; Yes, this all boils down to a fairly central question: do we b=
uild the same thing for HTTP/2 and HTTP/3?&nbsp; I'm not sure that we want t=
hat.&nbsp; We need to do a whole bunch of stuff to use QUIC (BTW, Roberto ge=
ts to say "told you so" again here), but doing that same work for HTTP/2 doe=
sn't make sense unless the changes are minimal.&nbsp; I'm suggesting that th=
e changes aren't as minimal as the rough sketches of protocols we have might=
 imply.<br>
<br>
&nbsp; &nbsp; If you conclude that there are more functions needed, as I hav=
e done, then you might also conclude that building those same functions into=
 two different protocols is not desirable, especially when there are no bene=
fits to doing so for one of those protocols.<br>
<br>
&nbsp; &nbsp; -- <br>
&nbsp; &nbsp; Webtransport mailing list<br>
&nbsp; &nbsp; <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Web=
transport@ietf.org</a><br>
&nbsp; &nbsp; <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport"=
 rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/=
webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@ietf=
.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"norefe=
rrer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport</=
a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@ietf=
.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"norefe=
rrer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport</=
a><br>
</blockquote></div>
</blockquote></div></div>
</blockquote></div>
</blockquote></div>
<span>-- </span><br><span>Webtransport mailing list</span><br><span>Webtrans=
port@ietf.org</span><br><span>https://www.ietf.org/mailman/listinfo/webtrans=
port</span><br></div></blockquote></div></div></body></html>=

--Apple-Mail-6A880F41-721F-4355-9C6D-A5E5573766A6--


From nobody Thu Jun  3 15:25:47 2021
Return-Path: <mt@lowentropy.net>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id D70403A1CDC for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 15:25:44 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_BLOCKED=0.001, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=lowentropy.net header.b=IBSPTg5/; dkim=pass (2048-bit key) header.d=messagingengine.com header.b=rvigYieK
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id hwc1pBewyu4w for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 15:25:38 -0700 (PDT)
Received: from out2-smtp.messagingengine.com (out2-smtp.messagingengine.com [66.111.4.26]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 65AED3A1CDA for <webtransport@ietf.org>; Thu,  3 Jun 2021 15:25:38 -0700 (PDT)
Received: from compute4.internal (compute4.nyi.internal [10.202.2.44]) by mailout.nyi.internal (Postfix) with ESMTP id 816665C01D3; Thu,  3 Jun 2021 18:25:36 -0400 (EDT)
Received: from imap10 ([10.202.2.60]) by compute4.internal (MEProxy); Thu, 03 Jun 2021 18:25:36 -0400
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=lowentropy.net; h=mime-version:message-id:in-reply-to:references:date:from:to :cc:subject:content-type:content-transfer-encoding; s=fm2; bh=+K /ozPIHtNzHePUNdNAvBJQcMme32NHeIxgN7uyDbDs=; b=IBSPTg5/9VE7HcW88Y XZLOvMdKUe6GnG3jmFBbkfokSaRNlb4SPve5yc6HbagAvlSgpA9eeaownDh9gGRX PcXZLBmWnuycWmk2vnp52XzCeoWybVQcCgmHSU+6jeG/L7nHT2Hux9KVzw0KisE0 MRwgNN744tKEkRD/+SVtMtvk0U64Bd317ad8StulRgtSNwcMcYEg1i1OZs3WCIwt 8cAVBGOb1M94hk69qXtHWMQyPD2RIFoswFjhZ6gEoUBYIrBmMSIXPvgFBzElYSRJ 7gDEHptAcaiJo9a560830R5NuQYCm5SZL6Z7UA/gpGBtJLT56yS3fRq76Qd7znC7 Fv6Q==
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d= messagingengine.com; h=cc:content-transfer-encoding:content-type :date:from:in-reply-to:message-id:mime-version:references :subject:to:x-me-proxy:x-me-proxy:x-me-sender:x-me-sender :x-sasl-enc; s=fm3; bh=+K/ozPIHtNzHePUNdNAvBJQcMme32NHeIxgN7uyDb Ds=; b=rvigYieKWyhZWueHUdDwh6MsUNwXnEjQigbeH2e37FfnDtamWLPjI2WFe PvECiLBAV5x7xKcHPVPIafDfs91iQQBxNGlEO6x4VVTgbMVkN722EBPH3kRf/Sir U/SlYV118pe0W4Aq0vm7dyHWHTzcYQ0boUppTvW2Qca0nZyZdOsiFlIPHfZY/2sg 936ZIDdMcl8WxCL73NUTym2PwlwkxzhDokmDr3xDMVdOXJ3Abha2d2gH00qZRDLW m2+i00S8uvw9HOl1cYQbYmTLJbY4hXf1AR7LvjfTwDS5mVfOTKpse5+gVsDiyHD3 f2yDfuqfhquB6vEGot3AyRjW+YNxA==
X-ME-Sender: <xms:31a5YHs3r9p70DKYhJuwNm5hUMbbPg47MlXiup2eTe_dCE7r0_gs7w> <xme:31a5YIc9zZ8LMxFkM2AIMxco7bvYbhaayH2mpQPba_u45eoVmO5QJeFoSDWqmqo8l m2s1eCj0iMOhxgw5yU>
X-ME-Proxy-Cause: gggruggvucftvghtrhhoucdtuddrgeduledrfedttddgtdegucetufdoteggodetrfdotf fvucfrrhhofhhilhgvmecuhfgrshhtofgrihhlpdfqfgfvpdfurfetoffkrfgpnffqhgen uceurghilhhouhhtmecufedttdenucesvcftvggtihhpihgvnhhtshculddquddttddmne cujfgurhepofgfggfkjghffffhvffutgfgsehtqhertderreejnecuhfhrohhmpedfofgr rhhtihhnucfvhhhomhhsohhnfdcuoehmtheslhhofigvnhhtrhhophihrdhnvghtqeenuc ggtffrrghtthgvrhhnpeevgfelgfelieffhffgtdeutdelgedvkefhkedufedtteduheev ieffhfefffehhfenucffohhmrghinhephhhtthhpfehvvghrshhiohhnrdhimhdpghhith hhuhgsrdgtohhmpdhhthhtphdvughovghsnhhtmhgrkhgvshgvnhhsvghunhhlvghsshht hhgvtghhrghnghgvshgrrhgvmhhinhhimhgrlhdrihhmpdhivghtfhdrohhrghenucevlh hushhtvghrufhiiigvpedtnecurfgrrhgrmhepmhgrihhlfhhrohhmpehmtheslhhofigv nhhtrhhophihrdhnvght
X-ME-Proxy: <xmx:31a5YKzUfjEZOxUkh0VWw5NGjec19s51h86zVt5RuA3D9BF0NtvkDg> <xmx:31a5YGOBhoE4C_q6ld0PnOX-1TqAGnJMXWmsKoV0ifZVh7KgEhjWTA> <xmx:31a5YH_NC1w7jjaUv02mKUksik95hmuP_yqDwZSSz--woe1-9UVsmA> <xmx:4Fa5YJJrTYOxI5HmAOomEWBUda_stFNUp8XRJoo-aTwm_ps78J4pjA>
Received: by mailuser.nyi.internal (Postfix, from userid 501) id BDAA14E02EB; Thu,  3 Jun 2021 18:25:35 -0400 (EDT)
X-Mailer: MessagingEngine.com Webmail Interface
User-Agent: Cyrus-JMAP/3.5.0-alpha0-519-g27a961944e-fm-20210531.001-g27a96194
Mime-Version: 1.0
Message-Id: <fa051e3c-7075-4bf8-af39-e3692d3360c5@www.fastmail.com>
In-Reply-To: <CAKcm_gMF3zpGP92tAv7MeKdFR+=Vc6uvGJ2PGFjTgxFWv9eJ7w@mail.gmail.com>
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com> <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com> <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com> <CAKcm_gNnQM5i7hR3V48ETnKX=WiizK7Z27w=cb3N_qLGdoB0LA@mail.gmail.com> <CAPDSy+5V8YhXVjCw1U4KcCcC+pzXZ7dBPJoYNDojEdACfnrO2Q@mail.gmail.com> <CAKcm_gMF3zpGP92tAv7MeKdFR+=Vc6uvGJ2PGFjTgxFWv9eJ7w@mail.gmail.com>
Date: Fri, 04 Jun 2021 08:25:16 +1000
From: "Martin Thomson" <mt@lowentropy.net>
To: "Ian Swett" <ianswett@google.com>, "David Schinazi" <dschinazi.ietf@gmail.com>
Cc: "Alan Frindell" <afrind@fb.com>, "Victor Vasiliev" <vasilvv@google.com>, WebTransport <webtransport@ietf.org>
Content-Type: text/plain;charset=utf-8
Content-Transfer-Encoding: quoted-printable
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/oi4YQpAOcAu3kKHz2Ao43p_aFz0>
Subject: Re: [Webtransport]  =?utf-8?q?Confirming_consensus_calls_from_today?= =?utf-8?q?=27s_WebTransport_interim?=
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Thu, 03 Jun 2021 22:25:45 -0000

One advantage of not using HTTP/2 streams directly is that you can ship =
something on top of HTTP/1.1.  And that might make deployment easier, no=
t so much because implementing HTTP/1.1 is easier (that's what people be=
lieve, but there are enough traps for the unwary that it might not be a =
true win), but more because it makes the resource allocation story simpl=
er if your WebTransport session takes up the entire connection.

I know that this is probably a little too flippant, but I don't see why =
we can't "just" drop the necessary QUIC frames into the request body.  S=
TREAM, DATAGRAM, MAX_STREAM_DATA, MAX_DATA, and MAX_STREAMS might be suf=
ficient.  Maybe PADDING and even PING.  It is not quite THAT trivial, as=
 you need to engage with prioritization and some other nasty things, but=
 it would still create a more portable and overall simpler design.

On Fri, Jun 4, 2021, at 02:40, Ian Swett wrote:
> Thanks David,
>=20
> I was thinking of corporate networks with MITMs that inspect HTTP 1.1=20=

> in a limited way, but it sounds like those may not be very common, so=20=

> there's no way to make WebTransport work anyway.
>=20
> If so, making WebTransport HTTP/2 specific is fine with me.
>=20
> I still think apps that want to work close to 100% of the time won't b=
e=20
> able to depend upon WebTransport always working, but maybe that's just=
=20
> the state of the internet.
>=20
> Ian
>=20
> On Thu, Jun 3, 2021 at 12:28 PM David Schinazi <dschinazi.ietf@gmail.c=
om> wrote:
> > Can you provide an example of a network that would benefit from this=
 scope increase? I've yet to see a network that MITMs TLS but allows all=
 HTTP/1.1 through unmodified. The ones that MITM and downgrade to h1 gen=
erally do it because they then inspect and/or modify the request and res=
ponse, so getting something with WebTransport semantics to work there at=
 all isn't guaranteed to be possible. Perhaps a good step forward is for=
 us to build a version of WebTransport over h2 and then look at deployme=
nt data and feedback to decide if another version is needed.
> >=20
> > David
> >=20
> > On Thu, Jun 3, 2021 at 8:56 AM Ian Swett <ianswett@google.com> wrote=
:
> >>=20
> >>=20
> >> On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <dschinazi.ietf@gmai=
l.com> wrote:
> >>> Hi Ian,
> >>>=20
> >>> Can you elaborate on what use-case would be solved by having WebTr=
ansport support HTTP/1 ?
> >>> - We know there are networks that block QUIC and/or UDP, those wil=
l be well served by HTTP/2.
> >>> - We know there are networks that MITM TLS then force a transparen=
t HTTP/1 proxy, those won't allow CONNECT nor any other form of WebTrans=
port anyway.
> >>> I'm failing to see a use-case where WebTransport over HTTP/1 would=
 be useful. Can you provide an example?
> >>=20
> >> I think there's evidence that those two sets of networks are very s=
imilar today, and over time I expect the set of networks that block QUIC=
 but don't block HTTP/2 to be tiny(ie: <1%).
> >>=20
> >> But I could be wrong.
> >>>=20
> >>> Thanks,
> >>> David
> >>>=20
> >>> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D40google.com@=
dmarc.ietf.org> wrote:
> >>>> I definitely agree with 1a, with the caveat that the TCP version =
is lower priority than the HTTP/3 version.
> >>>>=20
> >>>> I'm slightly less concerned about the stream state machine than M=
artin, but I think it's definitely a real concern and I think it makes s=
ense to have a solution fairly well fleshed out before adopting the dire=
ction and document.
> >>>>=20
> >>>> I'm slightly more concerned about creating a TCP fallback that do=
esn't work over HTTP/1.  If we're building a fallback, it seems like the=
se people shouldn't be excluded, since that means developers will have t=
o build and maintain a 3rd fallback that's not WebTransport based.  If t=
he dream is to be able to develop apps for WebTransport and not have a n=
on-WebTransport fallback, I don't see how we accomplish that without an =
option over HTTP/1?
> >>>>=20
> >>>> Possibly this is a forcing function to convince customers to upgr=
ade their middleboxes to ones which support HTTP/2 and HTTP/3, but if so=
 I'd like to call that out explicitly.
> >>>>=20
> >>>> Or maybe the physical office with its corp middleboxes is dead an=
d my concern is an unfounded anachronism?
> >>>>=20
> >>>> Ian
> >>>>=20
> >>>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D40fb.com@=
dmarc.ietf.org> wrote:
> >>>>> I think the fundamental discussion here is about whether we want=
 to modify/extend the HTTP/2 state machine to handle WebTransport stream=
s, or whether we should build an entirely separate stream state machine =
for them, and leave the H2 machinery untouched.
> >>>>>=20
> >>>>> Martin may be right that there be dragons down the path of modif=
ying the H2 state machine.  The alternative strikes me as an awful lot o=
f duplication -- it's close to implementing H2 again over a single HTTP =
stream, minus header processing.  That said, I think either approach can=
 be made to work, we just need to choose the tradeoffs we want. =20
> >>>>>=20
> >>>>> I am remiss in not addressing the feedback regarding unidirectio=
nal stream resets.  I think that it got lost during the bizarre hours of=
 the last IETF.  I filed an issue in the draft repo to track it (https:/=
/github.com/ekinnear/draft-webtransport-http2/issues/17).  I think the r=
ight solution is to keep the unidirectional reset functionality, and imp=
lement it in the H2 draft using either new frames or messages, depending=
 on the direction we take regarding the state machine.
> >>>>>=20
> >>>>> During the interim, we expressed our flexibility to adapt the de=
sign here as the H3 design evolves (eg: stream limits), new tools become=
 available (eg: CAPSULE) and as we gather implementation experience.
> >>>>>=20
> >>>>> While I think it is important we address this design decision as=
 soon as possible, I don't know that it needs to block adoption.=20
> >>>>>=20
> >>>>> Thanks
> >>>>>=20
> >>>>> -Alan
> >>>>>=20
> >>>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin =
Thomson" <webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> =
wrote:
> >>>>>=20
> >>>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
> >>>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <mt@lowentr=
opy.net> wrote:
> >>>>>     > > I'm opposed to adoption of this document.  I don't belie=
ve that it can be made so due to some fundamental differences in how HTT=
P/2 and QUIC stream states.
> >>>>>     > >=20
> >>>>>     > > If this problem was acknowledged, then I have every conf=
idence that the people listed as authors are able to find a solution.  B=
ut generally adoption is about the document more than the people.  Adopt=
ion means agreeing that the document contains approximately the right sh=
ape for a solution.  I can't agree with that in this case.
> >>>>>     >=20
> >>>>>     > Hi Martin,
> >>>>>     >=20
> >>>>>     > Could you elaborate on the specific differences between HT=
TP/2 and QUIC=20
> >>>>>     > stream state machines that you are concerned about?  The o=
nly one I am=20
> >>>>>     > aware of is unidirectional versus bidirectional reset, and=
 last time we=20
> >>>>>     > discussed this, the easiest solution was to make HTTP/3 be=
have like=20
> >>>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft woul=
d do to=20
> >>>>>     > address that.
> >>>>>=20
> >>>>>     That's the one I'm most concerned about, but there is also t=
he whole mess around inventing semantics for unidirectional streams that=
 needs to be worked into the state machines.
> >>>>>=20
> >>>>>     For reset, sure, you could make the HTTP/3 draft worse to de=
al with the problem.  Why would you want that?  QUIC has a much better s=
et of semantics here than HTTP/2.
> >>>>>=20
> >>>>>     There is also the stream state management disconnect for ser=
ver-initiated streams, which currently is only used for server push.  Th=
e draft currently doesn't say anywhere near as much as it would need to =
in order to specify that part.
> >>>>>=20
> >>>>>     And managing allocation of stream limits is a little scant o=
n detail in the same way.  I thought we wanted independent limits, which=
 means defining new accounting systems.  (Yes, this is a criticism of th=
e HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting somethin=
g in exchange.)
> >>>>>=20
> >>>>>     Why bother with all of that when you can just import QUIC fr=
ame semantics and then serialize them on the same stream?
> >>>>>=20
> >>>>>     Yes, this all boils down to a fairly central question: do we=
 build the same thing for HTTP/2 and HTTP/3?  I'm not sure that we want =
that.  We need to do a whole bunch of stuff to use QUIC (BTW, Roberto ge=
ts to say "told you so" again here), but doing that same work for HTTP/2=
 doesn't make sense unless the changes are minimal.  I'm suggesting that=
 the changes aren't as minimal as the rough sketches of protocols we hav=
e might imply.
> >>>>>=20
> >>>>>     If you conclude that there are more functions needed, as I h=
ave done, then you might also conclude that building those same function=
s into two different protocols is not desirable, especially when there a=
re no benefits to doing so for one of those protocols.
> >>>>>=20
> >>>>>     --=20
> >>>>>     Webtransport mailing list
> >>>>>     Webtransport@ietf.org
> >>>>>     https://www.ietf.org/mailman/listinfo/webtransport=20
> >>>>>=20
> >>>>> --=20
> >>>>> Webtransport mailing list
> >>>>> Webtransport@ietf.org
> >>>>> https://www.ietf.org/mailman/listinfo/webtransport
> >>>> --=20
> >>>> Webtransport mailing list
> >>>> Webtransport@ietf.org
> >>>> https://www.ietf.org/mailman/listinfo/webtransport


From nobody Thu Jun  3 17:20:09 2021
Return-Path: <prvs=57895d47b0=afrind@fb.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id B1A6D3A205F for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 17:20:05 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -3.495
X-Spam-Level: 
X-Spam-Status: No, score=-3.495 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.698, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_LOW=-0.7, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (1024-bit key) header.d=fb.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id seyef-lriKV0 for <webtransport@ietfa.amsl.com>; Thu,  3 Jun 2021 17:20:01 -0700 (PDT)
Received: from mx0a-00082601.pphosted.com (mx0b-00082601.pphosted.com [67.231.153.30]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id C9CE13A205E for <webtransport@ietf.org>; Thu,  3 Jun 2021 17:20:00 -0700 (PDT)
Received: from pps.filterd (m0001303.ppops.net [127.0.0.1]) by m0001303.ppops.net (8.16.0.43/8.16.0.43) with SMTP id 1540IvoP028464; Thu, 3 Jun 2021 17:19:57 -0700
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=fb.com; h=from : to : cc : subject : date : message-id : references : in-reply-to : content-type : content-id : content-transfer-encoding : mime-version; s=facebook; bh=cwv9rUk8tvDGiEze2OE38nkQmOEIimWRQ7UWoKzwYAQ=; b=LP839KNPifBlyUvmGzPrM7ThPqof6HFL3NL/M4iiVq5hMfZDUIlCqLipJo7bFHruU9H6 3+zJONFyrnklcJMUZaZZj/wJ0nzl9fK0yG23++Kp8jzKycob/JaPhbPEDbn66B5zyrut L1xM/EljFcrEVysp4oJe00T0ktou/Chf9as= 
Received: from mail.thefacebook.com ([163.114.132.120]) by m0001303.ppops.net with ESMTP id 38xkj6y8wj-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128 verify=NOT); Thu, 03 Jun 2021 17:19:56 -0700
Received: from NAM10-BN7-obe.outbound.protection.outlook.com (100.104.98.9) by o365-in.thefacebook.com (100.104.94.197) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2176.2; Thu, 3 Jun 2021 17:19:55 -0700
ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=g8x+3FHPSG+SJZiUOG6YviUaK4mgS0OjDgiLdWJYVIDLpbdxRL4C92PghiV3k3sWgg8FLg+z4eJ2mTy9gaUwn43kaIzmxv0OTgMk7OthNXYjy0WcOJFe6HHVDPezcsPB10VQOM6YL/VfxtWG1AVydjX61z87UZgnN8GjfAxO0JljCX/ZrGQlsrU3w9Wrt7JMlU80zb6kVWbxAz39ZFr7lQInW1Q1eTPqvqVuGLXUUZMeGB1/rNIF1L+9DzkOQ0I3bwhITHi58emmeQI5iAUx0/bCD0TOcd0CP6evINgIQN3+yv1jWgcM+2wj83r5XANVbDhPjbjS3FNpXhjjCvFDcQ==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com;  s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=QOdQbhXmAa7BwqJ1DudUzwKeJCyUrmSazCF4lwvzxf8=; b=WZnNIdp47l+pxGvz4jswd6+ZtKnFn8tg2cwnHNZcSvhPBVfhWo+b5EuLsWgOeqAmclCIdLPWLGae2JUfNwOsc+PvxohLwFQrk6rXf9sGa7R6AnNgJzGQwpCeK8gnCusNKd4YH1zOWtsAlRbqu3PaeRYnarMn4v2iMyjchK5ysjCKLSditDSu8MXz7TNvIuxO/CCuit5Lr27MaUGCIItrxT6eRteAViCneXwmHsNeW8fWNeACSU3D7ivD5RJkAdHMZw7TD+zcj2ixG8q8RlsoD30LyHzm4GbyL0+Fz9CQBhQY31LNhOVCb2zcEpM1g0Hh/53+GI1tL5/hEkuWnj+M8g==
ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=fb.com; dmarc=pass action=none header.from=fb.com; dkim=pass header.d=fb.com; arc=none
Received: from MW3PR15MB3772.namprd15.prod.outlook.com (2603:10b6:303:4c::14) by MWHPR15MB1534.namprd15.prod.outlook.com (2603:10b6:300:bf::20) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4173.26; Fri, 4 Jun 2021 00:19:54 +0000
Received: from MW3PR15MB3772.namprd15.prod.outlook.com ([fe80::b569:872e:690d:d455]) by MW3PR15MB3772.namprd15.prod.outlook.com ([fe80::b569:872e:690d:d455%7]) with mapi id 15.20.4195.023; Fri, 4 Jun 2021 00:19:53 +0000
From: Alan Frindell <afrind@fb.com>
To: Martin Thomson <mt@lowentropy.net>, Ian Swett <ianswett@google.com>, "David Schinazi" <dschinazi.ietf@gmail.com>
CC: Victor Vasiliev <vasilvv@google.com>, WebTransport <webtransport@ietf.org>
Thread-Topic: [Webtransport] Confirming consensus calls from today's WebTransport interim
Thread-Index: AQHXWIB4aU8iL3g6o0WCC41EQ6vMvKsCbAiAgAAE7QCAAAj8AIAAA2QAgABgUQD//6qugA==
Date: Fri, 4 Jun 2021 00:19:53 +0000
Message-ID: <7A74149F-B394-43FA-A4D0-BC5A85F2409F@fb.com>
References: <CAPDSy+5s-D7v3jxFmWKD5N-=sVcM_v8ZmB1iVNjZ1ZuHN5n=OQ@mail.gmail.com> <cce29e58-da55-4cc1-b760-4ee1c849c035@www.fastmail.com> <CAAZdMad9hZMutYAWQOO=T+UtV9tuMreHy707JO6qN40k2SVmMQ@mail.gmail.com> <69e2d8c9-b532-4ce3-b558-9fdd11e5fb1a@www.fastmail.com> <304A9399-D211-4463-9245-C2437E318FC3@fb.com> <CAKcm_gOLTXtjf_jQWZ12sKe4L-e5J3jA+Zn5zjd7UJrVYQ1f8w@mail.gmail.com> <CAPDSy+7hBj8EAJcydCGb5iVh3XxFoaBLzKgb=4X7myP-2-ZEZA@mail.gmail.com> <CAKcm_gNnQM5i7hR3V48ETnKX=WiizK7Z27w=cb3N_qLGdoB0LA@mail.gmail.com> <CAPDSy+5V8YhXVjCw1U4KcCcC+pzXZ7dBPJoYNDojEdACfnrO2Q@mail.gmail.com> <CAKcm_gMF3zpGP92tAv7MeKdFR+=Vc6uvGJ2PGFjTgxFWv9eJ7w@mail.gmail.com> <fa051e3c-7075-4bf8-af39-e3692d3360c5@www.fastmail.com>
In-Reply-To: <fa051e3c-7075-4bf8-af39-e3692d3360c5@www.fastmail.com>
Accept-Language: en-US
Content-Language: en-US
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
user-agent: Microsoft-MacOutlook/16.49.21050901
authentication-results: lowentropy.net; dkim=none (message not signed) header.d=none;lowentropy.net; dmarc=none action=none header.from=fb.com;
x-originating-ip: [2620:10d:c090:400::5:e3d]
x-ms-publictraffictype: Email
x-ms-office365-filtering-correlation-id: 8b65aca8-c75c-4365-f76f-08d926ee79eb
x-ms-traffictypediagnostic: MWHPR15MB1534:
x-microsoft-antispam-prvs: <MWHPR15MB1534B4A546C63A39140E7662A73B9@MWHPR15MB1534.namprd15.prod.outlook.com>
x-fb-source: Internal
x-ms-oob-tlc-oobclassifiers: OLM:7219;
x-ms-exchange-senderadcheck: 1
x-microsoft-antispam: BCL:0;
x-microsoft-antispam-message-info: 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
x-forefront-antispam-report: CIP:255.255.255.255; CTRY:; LANG:en; SCL:1; SRV:;  IPV:NLI; SFV:NSPM; H:MW3PR15MB3772.namprd15.prod.outlook.com; PTR:; CAT:NONE;  SFS:(4636009)(136003)(346002)(39860400002)(376002)(366004)(396003)(4326008)(38100700002)(66446008)(122000001)(478600001)(91956017)(64756008)(316002)(86362001)(5660300002)(6512007)(110136005)(54906003)(6506007)(6486002)(66946007)(66574015)(36756003)(76116006)(66556008)(71200400001)(66476007)(53546011)(186003)(8936002)(8676002)(2616005)(2906002)(33656002)(966005)(83380400001)(45980500001); DIR:OUT; SFP:1102; 
x-ms-exchange-antispam-messagedata: =?utf-8?B?ZnNOVzRwTVh5WkJBanZGUDdrZ2tvbmRyZGpzWjVQRlF3Nmcwa2hWQjdpY2N1?= =?utf-8?B?L0RQN0J0Nmd6ajNFREpPTmpkMnNqZWwyVVdJMWh4V3NnbXR3T2t6OTJJVVk1?= =?utf-8?B?UWdMd29vV3BJUDhpV0xoaFh0RlVydEtXS3BRTzVHM0VCQUVKbUtzRVpORzd5?= =?utf-8?B?OXg0OENMZVNaeEFkdVo5RkxmREQ2aDVhMnozK1hOOGU0NjhWazZrbnAwRi9K?= =?utf-8?B?N043VFpWLzNQWG10SkFKS214V3hLUWVIMlNHdnRGREhLN0E4ME1Fa1BUWG1k?= =?utf-8?B?dkZVM2ZlVnNHZ3BwcTRjMFoydUJjZFQ4RGdTUGhuTGVqNHRmTlRkbDQ3Rldz?= =?utf-8?B?WW1LWTB3QUxIQkY2WFNzSjZFOFN6cmxOYVBteU1zWG5XODBPeW1mUDdGbDh3?= =?utf-8?B?dzdxQkdYcmlwTitCSVJzSFhkUTY2aUkyL0ZQWmRVN0JSWnJnQWxVSmpFckpl?= =?utf-8?B?RDNlVzlqR2QwN1pETFBuQ1FWQmZaVEIrQllzaGV5QS96OTRKOHRVNnJKNk50?= =?utf-8?B?MG0xc1RlUCsrRzUraXlIOXJzQW9HQVEwUnp0cXZGem1vM2VPem5wNTZUa3NQ?= =?utf-8?B?NFI2VktLdFVNdnJGUm5yR0prT0hMTTh3blVlT3k3Qm4wSUtFU3NYZXRNU2RE?= =?utf-8?B?UUQwS2swZ2ZJVVp1cURKRWpWTVRMT2RUa2xsb2dENFIya0NKQkcydGRwR0wr?= =?utf-8?B?YXB4MjRoUVQ2UjRzVHcvWkl4cDRMbFNiNzhMU0NQK1p0YVEzY0hnQWhkS2Q5?= =?utf-8?B?ZVBNWWpjWm5LM2JlT1lSd1RTRFBIMFErWWJsOE84L2dYWENpRnFmVmpYV09p?= =?utf-8?B?bDNFVG9rVU54aTMrYTNyRmxTcS9LR2lSNHoycG5zWS9ITjBnM2Nwc2s0amV3?= =?utf-8?B?L3NNbUxpR05Jby9mUlNUN0xJVDErWWtjSW9KN0xob1lkUWdlMmVuL2dGOTRR?= =?utf-8?B?S2xGeWNZWE9VWVpjN1RuSk80cXhINDdLTnpyNG5pRDlqQjAwZll4b21Wcnoz?= =?utf-8?B?WHVZRWhDakpQUjltSWRvNmt6VlJkUEZvZXRZS05ITGFxcWh4VGFkdis0NmtW?= =?utf-8?B?YXVUcWJhRjZxOVh0UGRxWUhSUlNlMXNtVVYxeGZPWW9ZbUY4K0h5TTdRVVAw?= =?utf-8?B?RlJya3o3ZXk3UGVIN0RnUThSbUJjTS9ic0dyUnhyRlp6R3F0WlJMQ1FuekNF?= =?utf-8?B?K05SN0I0dmtTdWUwSTFZbkpmaDhOYjF4S01YeTVSbkkrbmZ2WXdxellyM2U1?= =?utf-8?B?YUNDTXNhRHhUQjNCYXpNUTRGWGhiU25vT1ltc0NQN3VpTTU1WUlyQVNrdnIv?= =?utf-8?B?dlpQK3lkQ3ZuR0N1V0IzdW5NdmZTR25nbTR5bVEreU1EYjBVL09WS2RabWxW?= =?utf-8?B?VzVkU0s5dForeXhhdVd3OVY4MTNQQ2drVkJMZXBLN1RXOUFJaSs5VUJMVlBY?= =?utf-8?B?YW9rcnZjckNXVnNxU1k0VFF6N1VFTXl0Z2JRQTRqbkNoeFQwVFc2N2Y2QWdL?= =?utf-8?B?SUJPNUFQcitYKzlMNWJieW1jUzBvS202M1lGVkI2NVI3em5Daldtc1hOY3VC?= =?utf-8?B?dTMxa3NNb1pQK2NNbUFHVGN6NWpHRmxDd3pxVVowYWQ3eStOUjFRcSt5RnUz?= =?utf-8?B?Um52K0llZG96SEEzSkRZNS90VitYV3VuWmhsL1psb1o0Z2RnY2w0bHhOTWFE?= =?utf-8?B?K2RSbzFWSDN0NmsweDJHMU1rejFEd1JxTnNOODRnaU5yUnlmYmdiNzdUWTRI?= =?utf-8?B?cHVIWkIrSU8zNEx0bE9PbXpQL20rNnc3dTFWT29ROGhIRXFYdEZzMXo4eVcz?= =?utf-8?Q?jIphxGHO6z+PcDSpAced0lEoswdCpmb9v6b3g=3D?=
x-ms-exchange-transport-forked: True
Content-Type: text/plain; charset="utf-8"
Content-ID: <99381DA5FBB9C44DAB976B5CBF1E8460@namprd15.prod.outlook.com>
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: MW3PR15MB3772.namprd15.prod.outlook.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 8b65aca8-c75c-4365-f76f-08d926ee79eb
X-MS-Exchange-CrossTenant-originalarrivaltime: 04 Jun 2021 00:19:53.7343 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 8ae927fe-1255-47a7-a2af-5f3a069daaa2
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-CrossTenant-userprincipalname: zzycWYC/Q+pRD91gryHW5OI17EivNaOLclM6+BdNEx+v5S79qkkwYFKbUl9+kFsu
X-MS-Exchange-Transport-CrossTenantHeadersStamped: MWHPR15MB1534
X-OriginatorOrg: fb.com
X-Proofpoint-GUID: SBbdBcHLx4I2oZbXYtzcEAtS9maHeL97
X-Proofpoint-ORIG-GUID: SBbdBcHLx4I2oZbXYtzcEAtS9maHeL97
Content-Transfer-Encoding: base64
X-Proofpoint-UnRewURL: 3 URL's were un-rewritten
MIME-Version: 1.0
X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:6.0.391, 18.0.761 definitions=2021-06-03_15:2021-06-03, 2021-06-03 signatures=0
X-Proofpoint-Spam-Details: rule=fb_default_notspam policy=fb_default score=0 lowpriorityscore=0 impostorscore=0 mlxlogscore=999 priorityscore=1501 mlxscore=0 phishscore=0 clxscore=1011 suspectscore=0 spamscore=0 malwarescore=0 adultscore=0 bulkscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2104190000 definitions=main-2106040000
X-FB-Internal: deliver
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/cKbV13CjEf34if57Twc8scLyzCg>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 04 Jun 2021 00:20:06 -0000
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From nobody Fri Jun  4 06:54:35 2021
Return-Path: <ianswett@google.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 590B33A11F6 for <webtransport@ietfa.amsl.com>; Fri,  4 Jun 2021 06:54:33 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -17.599
X-Spam-Level: 
X-Spam-Status: No, score=-17.599 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIMWL_WL_MED=-0.001, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, ENV_AND_HDR_SPF_MATCH=-0.5, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001, USER_IN_DEF_DKIM_WL=-7.5, USER_IN_DEF_SPF_WL=-7.5] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=google.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id PZLJM1jusJg4 for <webtransport@ietfa.amsl.com>; Fri,  4 Jun 2021 06:54:28 -0700 (PDT)
Received: from mail-wr1-x436.google.com (mail-wr1-x436.google.com [IPv6:2a00:1450:4864:20::436]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 0A2543A11E8 for <webtransport@ietf.org>; Fri,  4 Jun 2021 06:54:27 -0700 (PDT)
Received: by mail-wr1-x436.google.com with SMTP id c5so9377677wrq.9 for <webtransport@ietf.org>; Fri, 04 Jun 2021 06:54:27 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=odDHgFCjbagUYvm1wOPuVO069+TBylKo3+oAye3hxFk=; b=ZX3QtMI7nrgpleIcRPbcHEyO92yRv4oUiIwPAcRw2M3ygr05OmN3P6gq7k+CwsW3yX A2DmIcgbXi0WWbu66eSnCovIBuDalT3MaIkN+mMpGc41tuyq2haG3yaxs7CnYT0OsEWQ lAycLDfw9TLIXfH0EUj8x4bz3lcaLMomdQ+HEo1/ttP8tkJwk0j5oXqv4En373Q3Z35w D2Uo4+RU9rY+pJI4h0XKTpng4Uv2QcCSSBi4gvFS1I9OQ/153bJXPT1by9bsz5Crdujf O3Qt84nnx8Mp/SuPNv4MmN6pUnxwY3RV/sY+AmhigTWTtn+e5aMtbm81Z5ozJp0nmV4b ZDdQ==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=odDHgFCjbagUYvm1wOPuVO069+TBylKo3+oAye3hxFk=; b=OvnczZpEfppWHQi0Oga+16lDw39j7mq2wOdiXbf3p6vHFw3LFcpGX1UJbxfmkyYIWp 7QQUSxoNueBpUscz2Diuw5q5LnH0eddchG20Yiqkmu0Xl+JBjW1GRLsarWvZb2erEX6K m+wC2ZooJt5/KrpbEEE2QW8lv/R4gL+cqNCShcmsjPf7XfAZoTlRVUrSKafAqqDmBkzU 827/lKw0Ljxb0LN3CywInFizRsk3au6JJRbFES3pxDVBOOpn8KabxQtSh8fHTEddwu8m q52QTx6fm3MPhN6sKhWOtWZh9u0szIMxT1BiHxYWtatCBOGKLyldtmoQ+RLsRzUZfP6/ Q12Q==
X-Gm-Message-State: AOAM533+CKqooVq2GmLtmr+MaeYEKv03k4nCtDkyWIBNuaFVxuUuDWSY 7alJLNWTEkRJUwEGgApYOUt4XrHWPi2+scaZlZTf5g==
X-Google-Smtp-Source: ABdhPJwYeS5NhSJz4HZMGybDsDk60W7npIe7z32n3gftJkXuhtkIDH5SxH6QUnHHqpTKQ50yH5KKY3JrOOFs/RkjnIg=
X-Received: by 2002:a5d:474d:: with SMTP id o13mr3985321wrs.176.1622814865068;  Fri, 04 Jun 2021 06:54:25 -0700 (PDT)
MIME-Version: 1.0
References: <7D4F22BF-6910-4E45-BC27-F395461B603C@apple.com>
In-Reply-To: <7D4F22BF-6910-4E45-BC27-F395461B603C@apple.com>
From: Ian Swett <ianswett@google.com>
Date: Fri, 4 Jun 2021 09:54:12 -0400
Message-ID: <CAKcm_gOywB_P9pRM3fUAdzJ+uh9+Snrwyz-0_zP+jXFhWNmrEw@mail.gmail.com>
To: Eric Kinnear <ekinnear=40apple.com@dmarc.ietf.org>
Cc: David Schinazi <dschinazi.ietf@gmail.com>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>, Alan Frindell <afrind@fb.com>,  Martin Thomson <mt@lowentropy.net>
Content-Type: multipart/alternative; boundary="00000000000091f34705c3f107db"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/gaAT06M0QDFD3ABStR_NG41v_DM>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Fri, 04 Jun 2021 13:54:33 -0000

--00000000000091f34705c3f107db
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

I've always thought QUIC over HTTP(or TCP) would be cool, but this is the
first time I've heard a compelling use case.  Definitely worth considering.

On Thu, Jun 3, 2021 at 2:17 PM Eric Kinnear <ekinnear=3D
40apple.com@dmarc.ietf.org> wrote:

> =EF=BB=BFI think the question of HTTP/1.1 is an interesting one: we=E2=80=
=99ve so far been
> operating with the view that falling back from QUIC to TLS/TCP might enab=
le
> significantly more folks to participate, but that the delta between
> HTTP/1.1 and HTTP/2 would not be as significant.
>
> I don=E2=80=99t have massively strong opinions in terms of how we go abou=
t
> addressing that, beyond the earlier concerns about additional effort
> required to cover HTTP/1.1 vs. expected gain, but if the expected gain go=
es
> up, then perhaps that changes things sufficiently.
>
> It seems that some of that thinking has come from the observations made
> about _who_ is in control of the versioning; if your network operator doe=
s
> not allow QUIC then that=E2=80=99s less under your control than whether H=
TTP/2 is
> supported (although, as Ian mentioned, there=E2=80=99s a population with
> middleboxes that force you down to HTTP/1.1, which might or might not
> support CONNECT anyways).
>
> Would it be possible to gather some data about the relative size of these
> different populations? I=E2=80=99m not sure there=E2=80=99s anyone in a g=
reat position to
> do that, at least for CONNECT specifically if it=E2=80=99s not already
> possible/in-use, but perhaps data about other H1 vs. H2 vs. H3 usage woul=
d
> be interesting.
>

Google has server-side data of "For people in an HTTP/3 enabled group, what
fraction of people end up using HTTP/3, HTTP/2, and HTTP/1" and we can
breakdown by OS/etc.  I'm happy to share but I'd need approvals, so if I
know exactly what data people would like to see, I can ensure it's all
approved at once(ie: a single slide deck).

This seems like a fun topic for maprg, but I'm not sure it's worth waiting
that long?

It would be ideal if different people could share their data, since any
single set of users may not be completely representative(ie: one could
block a set of domains entirely, so none of the protocols worked).


>
> We know that, at least on iOS, HTTP/2 is in use in a higher percentage fo=
r
> web browsing use cases than for other application use cases.
>
> On Jun 3, 2021, at 12:41 PM, Ian Swett <ianswett=3D
> 40google.com@dmarc.ietf.org> wrote:
>
> =EF=BB=BF
> Thanks David,
>
> I was thinking of corporate networks with MITMs that inspect HTTP 1.1 in =
a
> limited way, but it sounds like those may not be very common, so there's =
no
> way to make WebTransport work anyway.
>
> If so, making WebTransport HTTP/2 specific is fine with me.
>
> I still think apps that want to work close to 100% of the time won't be
> able to depend upon WebTransport always working, but maybe that's just th=
e
> state of the internet.
>
> Ian
>
> On Thu, Jun 3, 2021 at 12:28 PM David Schinazi <dschinazi.ietf@gmail.com>
> wrote:
>
>> Can you provide an example of a network that would benefit from this
>> scope increase? I've yet to see a network that MITMs TLS but allows all
>> HTTP/1.1 through unmodified. The ones that MITM and downgrade to h1
>> generally do it because they then inspect and/or modify the request and
>> response, so getting something with WebTransport semantics to work there=
 at
>> all isn't guaranteed to be possible. Perhaps a good step forward is for =
us
>> to build a version of WebTransport over h2 and then look at deployment d=
ata
>> and feedback to decide if another version is needed.
>>
>> David
>>
>> On Thu, Jun 3, 2021 at 8:56 AM Ian Swett <ianswett@google.com> wrote:
>>
>>>
>>>
>>> On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <dschinazi.ietf@gmail.co=
m>
>>> wrote:
>>>
>>>> Hi Ian,
>>>>
>>>> Can you elaborate on what use-case would be solved by having
>>>> WebTransport support HTTP/1 ?
>>>> - We know there are networks that block QUIC and/or UDP, those will be
>>>> well served by HTTP/2.
>>>> - We know there are networks that MITM TLS then force a transparent
>>>> HTTP/1 proxy, those won't allow CONNECT nor any other form of WebTrans=
port
>>>> anyway.
>>>> I'm failing to see a use-case where WebTransport over HTTP/1 would be
>>>> useful. Can you provide an example?
>>>>
>>>
>>> I think there's evidence that those two sets of networks are very
>>> similar today, and over time I expect the set of networks that block QU=
IC
>>> but don't block HTTP/2 to be tiny(ie: <1%).
>>>
>>> But I could be wrong.
>>>
>>>>
>>>> Thanks,
>>>> David
>>>>
>>>> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D
>>>> 40google.com@dmarc.ietf.org> wrote:
>>>>
>>>>> I definitely agree with 1a, with the caveat that the TCP version is
>>>>> lower priority than the HTTP/3 version.
>>>>>
>>>>> I'm slightly less concerned about the stream state machine than
>>>>> Martin, but I think it's definitely a real concern and I think it mak=
es
>>>>> sense to have a solution fairly well fleshed out before adopting the
>>>>> direction and document.
>>>>>
>>>>> I'm slightly more concerned about creating a TCP fallback that doesn'=
t
>>>>> work over HTTP/1.  If we're building a fallback, it seems like these =
people
>>>>> shouldn't be excluded, since that means developers will have to build=
 and
>>>>> maintain a 3rd fallback that's not WebTransport based.  If the dream =
is to
>>>>> be able to develop apps for WebTransport and not have a non-WebTransp=
ort
>>>>> fallback, I don't see how we accomplish that without an option over H=
TTP/1?
>>>>>
>>>>> Possibly this is a forcing function to convince customers to upgrade
>>>>> their middleboxes to ones which support HTTP/2 and HTTP/3, but if so =
I'd
>>>>> like to call that out explicitly.
>>>>>
>>>>> Or maybe the physical office with its corp middleboxes is dead and my
>>>>> concern is an unfounded anachronism?
>>>>>
>>>>> Ian
>>>>>
>>>>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
>>>>> 40fb.com@dmarc.ietf.org> wrote:
>>>>>
>>>>>> I think the fundamental discussion here is about whether we want to
>>>>>> modify/extend the HTTP/2 state machine to handle WebTransport stream=
s, or
>>>>>> whether we should build an entirely separate stream state machine fo=
r them,
>>>>>> and leave the H2 machinery untouched.
>>>>>>
>>>>>
> This sounds about right to me, if we think that the mapping of H3 onto H2
> is not how we=E2=80=99d want it to be, we=E2=80=99ve evolved it significa=
ntly so far, and I
> don=E2=80=99t see any reason why we can=E2=80=99t continue to do so.
>
>
>>>>>> Martin may be right that there be dragons down the path of modifying
>>>>>> the H2 state machine.  The alternative strikes me as an awful lot of
>>>>>> duplication -- it's close to implementing H2 again over a single HTT=
P
>>>>>> stream, minus header processing.  That said, I think either approach=
 can be
>>>>>> made to work, we just need to choose the tradeoffs we want.
>>>>>>
>>>>>> I am remiss in not addressing the feedback regarding unidirectional
>>>>>> stream resets.  I think that it got lost during the bizarre hours of=
 the
>>>>>> last IETF.  I filed an issue in the draft repo to track it (
>>>>>> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I
>>>>>> think the right solution is to keep the unidirectional reset functio=
nality,
>>>>>> and implement it in the H2 draft using either new frames or messages=
,
>>>>>> depending on the direction we take regarding the state machine.
>>>>>>
>>>>>> During the interim, we expressed our flexibility to adapt the design
>>>>>> here as the H3 design evolves (eg: stream limits), new tools become
>>>>>> available (eg: CAPSULE) and as we gather implementation experience.
>>>>>>
>>>>>> While I think it is important we address this design decision as soo=
n
>>>>>> as possible, I don't know that it needs to block adoption.
>>>>>>
>>>>>
> I=E2=80=99d agree, I support adoption (as an author), and with that I=E2=
=80=99d very much
> like to sit down and make sure we=E2=80=99re covering Martin=E2=80=99s fe=
edback here =E2=80=94 I
> know parts of it were also raised at 110 prior to the interim and we
> haven=E2=80=99t all synced up to make sure it=E2=80=99s sorted out. That =
might result in
> minor changes to the document, it might result in major changes to the
> document =E2=80=94 either is fine from my perspective.
>
> Thanks,
> Eric
>
>
>
>>>>>> Thanks
>>>>>>
>>>>>> -Alan
>>>>>>
>>>>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Thom=
son" <
>>>>>> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote:
>>>>>>
>>>>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>>>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <
>>>>>> mt@lowentropy.net> wrote:
>>>>>>     > > I'm opposed to adoption of this document.  I don't believe
>>>>>> that it can be made so due to some fundamental differences in how HT=
TP/2
>>>>>> and QUIC stream states.
>>>>>>     > >
>>>>>>     > > If this problem was acknowledged, then I have every
>>>>>> confidence that the people listed as authors are able to find a solu=
tion.
>>>>>> But generally adoption is about the document more than the people.
>>>>>> Adoption means agreeing that the document contains approximately the=
 right
>>>>>> shape for a solution.  I can't agree with that in this case.
>>>>>>     >
>>>>>>     > Hi Martin,
>>>>>>     >
>>>>>>     > Could you elaborate on the specific differences between HTTP/2
>>>>>> and QUIC
>>>>>>     > stream state machines that you are concerned about?  The only
>>>>>> one I am
>>>>>>     > aware of is unidirectional versus bidirectional reset, and las=
t
>>>>>> time we
>>>>>>     > discussed this, the easiest solution was to make HTTP/3 behave
>>>>>> like
>>>>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would do
>>>>>> to
>>>>>>     > address that.
>>>>>>
>>>>>>     That's the one I'm most concerned about, but there is also the
>>>>>> whole mess around inventing semantics for unidirectional streams tha=
t needs
>>>>>> to be worked into the state machines.
>>>>>>
>>>>>>     For reset, sure, you could make the HTTP/3 draft worse to deal
>>>>>> with the problem.  Why would you want that?  QUIC has a much better =
set of
>>>>>> semantics here than HTTP/2.
>>>>>>
>>>>>>     There is also the stream state management disconnect for
>>>>>> server-initiated streams, which currently is only used for server pu=
sh.
>>>>>> The draft currently doesn't say anywhere near as much as it would ne=
ed to
>>>>>> in order to specify that part.
>>>>>>
>>>>>>     And managing allocation of stream limits is a little scant on
>>>>>> detail in the same way.  I thought we wanted independent limits, whi=
ch
>>>>>> means defining new accounting systems.  (Yes, this is a criticism of=
 the
>>>>>> HTTP/3 draft also, but the additional effort required to use QUIC st=
reams
>>>>>> is entirely justified in that case, because you are getting somethin=
g in
>>>>>> exchange.)
>>>>>>
>>>>>>     Why bother with all of that when you can just import QUIC frame
>>>>>> semantics and then serialize them on the same stream?
>>>>>>
>>>>>>     Yes, this all boils down to a fairly central question: do we
>>>>>> build the same thing for HTTP/2 and HTTP/3?  I'm not sure that we wa=
nt
>>>>>> that.  We need to do a whole bunch of stuff to use QUIC (BTW, Robert=
o gets
>>>>>> to say "told you so" again here), but doing that same work for HTTP/=
2
>>>>>> doesn't make sense unless the changes are minimal.  I'm suggesting t=
hat the
>>>>>> changes aren't as minimal as the rough sketches of protocols we have=
 might
>>>>>> imply.
>>>>>>
>>>>>>     If you conclude that there are more functions needed, as I have
>>>>>> done, then you might also conclude that building those same function=
s into
>>>>>> two different protocols is not desirable, especially when there are =
no
>>>>>> benefits to doing so for one of those protocols.
>>>>>>
>>>>>>     --
>>>>>>     Webtransport mailing list
>>>>>>     Webtransport@ietf.org
>>>>>>     https://www.ietf.org/mailman/listinfo/webtransport
>>>>>>
>>>>>> --
>>>>>> Webtransport mailing list
>>>>>> Webtransport@ietf.org
>>>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>>>
>>>>> --
>>>>> Webtransport mailing list
>>>>> Webtransport@ietf.org
>>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>>
>>>> --
> Webtransport mailing list
> Webtransport@ietf.org
> https://www.ietf.org/mailman/listinfo/webtransport
>
>

--00000000000091f34705c3f107db
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>I&#39;ve always=C2=A0thought QUIC over HTTP(or TCP) w=
ould be cool, but this is the first time I&#39;ve heard a compelling use ca=
se.=C2=A0 Definitely worth considering.</div><br><div class=3D"gmail_quote"=
><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 2:17 PM Eric =
Kinnear &lt;ekinnear=3D<a href=3D"mailto:40apple.com@dmarc.ietf.org">40appl=
e.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quo=
te" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204=
);padding-left:1ex"><div dir=3D"auto"><div dir=3D"ltr">=EF=BB=BFI think the=
 question of HTTP/1.1 is an interesting one: we=E2=80=99ve so far been oper=
ating with the view that falling back from QUIC to TLS/TCP might enable sig=
nificantly more folks to participate, but that the delta between HTTP/1.1 a=
nd HTTP/2 would not be as significant.<div dir=3D"ltr"><div dir=3D"ltr"><br=
></div><div dir=3D"ltr">I don=E2=80=99t have massively strong opinions in t=
erms of how we go about addressing that, beyond the earlier concerns about =
additional effort required to cover HTTP/1.1 vs. expected gain, but if the =
expected gain goes up, then perhaps that changes things sufficiently.=C2=A0=
</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">It seems that some of tha=
t thinking has come from the observations made about _who_ is in control of=
 the versioning; if your network operator does not allow QUIC then that=E2=
=80=99s less under your control than whether HTTP/2 is supported (although,=
 as Ian mentioned, there=E2=80=99s a population with middleboxes that force=
 you down to HTTP/1.1, which might or might not support CONNECT anyways).=
=C2=A0</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">Would it be possibl=
e to gather some data about the relative size of these different population=
s? I=E2=80=99m not sure there=E2=80=99s anyone in a great position to do th=
at, at least for CONNECT specifically if it=E2=80=99s not already possible/=
in-use, but perhaps data about other H1 vs. H2 vs. H3 usage would be intere=
sting.</div></div></div></div></blockquote><div><br></div><div>Google has s=
erver-side data of &quot;For people in an HTTP/3 enabled group, what fracti=
on of people end up using HTTP/3, HTTP/2, and HTTP/1&quot; and we can break=
down by OS/etc.=C2=A0 I&#39;m happy to share but I&#39;d need approvals, so=
 if I know exactly what data people would like to see, I can ensure it&#39;=
s all approved at once(ie: a single slide deck).</div><div><br></div><div>T=
his seems like a fun topic for maprg, but I&#39;m not sure it&#39;s worth w=
aiting that long?</div><div><br></div><div>It would be ideal if different p=
eople could share their data, since any single set of users may not be comp=
letely representative(ie: one could block a set of domains entirely, so non=
e of the protocols worked).</div><div>=C2=A0</div><blockquote class=3D"gmai=
l_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,20=
4,204);padding-left:1ex"><div dir=3D"auto"><div dir=3D"ltr"><div dir=3D"ltr=
"><div dir=3D"ltr"><br></div><div dir=3D"ltr">We know that, at least on iOS=
, HTTP/2 is in use in a higher percentage for web browsing use cases than f=
or other application use cases.</div><div dir=3D"ltr"><br><blockquote type=
=3D"cite">On Jun 3, 2021, at 12:41 PM, Ian Swett &lt;ianswett=3D<a href=3D"=
mailto:40google.com@dmarc.ietf.org" target=3D"_blank">40google.com@dmarc.ie=
tf.org</a>&gt; wrote:<br><br></blockquote></div><blockquote type=3D"cite"><=
div dir=3D"ltr">=EF=BB=BF<div dir=3D"ltr">Thanks David,<div><br></div><div>=
I was thinking of corporate networks with MITMs that inspect HTTP 1.1 in a =
limited way, but it sounds like those may not be very common, so there&#39;=
s no way to make WebTransport work anyway.</div><div><br></div><div>If so, =
making WebTransport HTTP/2 specific is fine with me.</div><div><br></div><d=
iv>I still think apps that want to work close to 100% of the time won&#39;t=
 be able to depend upon WebTransport always working, but maybe that&#39;s j=
ust the state of the internet.</div><div><br></div><div>Ian</div></div><br>=
<div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Ju=
n 3, 2021 at 12:28 PM David Schinazi &lt;<a href=3D"mailto:dschinazi.ietf@g=
mail.com" target=3D"_blank">dschinazi.ietf@gmail.com</a>&gt; wrote:<br></di=
v><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr">Can yo=
u provide an example of a network that would benefit from this scope increa=
se? I&#39;ve yet to see a network that MITMs TLS but allows all HTTP/1.1 th=
rough unmodified. The ones that MITM and downgrade to h1 generally=C2=A0do =
it because they then inspect and/or modify the request and response, so get=
ting something with WebTransport semantics to work there at all isn&#39;t g=
uaranteed to be possible. Perhaps a good step forward is for us to build a =
version of WebTransport over h2 and then look at deployment data and feedba=
ck to decide if another version is needed.<div><br></div><div>David</div></=
div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On=
 Thu, Jun 3, 2021 at 8:56 AM Ian Swett &lt;<a href=3D"mailto:ianswett@googl=
e.com" target=3D"_blank">ianswett@google.com</a>&gt; wrote:<br></div><block=
quote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1=
px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"lt=
r"><br></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail=
_attr">On Thu, Jun 3, 2021 at 11:38 AM David Schinazi &lt;<a href=3D"mailto=
:dschinazi.ietf@gmail.com" target=3D"_blank">dschinazi.ietf@gmail.com</a>&g=
t; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0p=
x 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div d=
ir=3D"ltr">Hi Ian,<div><br></div><div>Can you elaborate on what use-case wo=
uld be solved by having WebTransport support HTTP/1 ?</div><div>- We know t=
here are networks that block QUIC and/or UDP, those will be well served by =
HTTP/2.</div><div>- We know there are networks that MITM TLS then force a t=
ransparent HTTP/1 proxy, those won&#39;t allow CONNECT nor any other form o=
f WebTransport anyway.</div><div>I&#39;m failing to see a use-case where We=
bTransport over HTTP/1 would be useful. Can you provide an example?</div></=
div></blockquote><div><br></div><div>I think there&#39;s evidence that thos=
e two sets of networks are very similar today, and over time I expect the s=
et of networks that block QUIC but don&#39;t block HTTP/2 to be tiny(ie: &l=
t;1%).</div><div><br></div><div>But I could be wrong.</div><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><br></div><div>Than=
ks,</div><div>David</div></div><br><div class=3D"gmail_quote"><div dir=3D"l=
tr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 6:57 AM Ian Swett &lt;iansw=
ett=3D<a href=3D"mailto:40google.com@dmarc.ietf.org" target=3D"_blank">40go=
ogle.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_=
quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,=
204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"></div>I definitely=
 agree with 1a, with the caveat that the=C2=A0TCP version is lower priority=
 than the HTTP/3 version.<div><br></div><div>I&#39;m slightly less concerne=
d about the stream state machine than Martin, but I think it&#39;s definite=
ly a real concern and I think it makes sense to have a solution fairly well=
 fleshed out before adopting the direction and document.</div><div><br></di=
v><div>I&#39;m slightly more concerned about creating=C2=A0a TCP fallback t=
hat doesn&#39;t work over HTTP/1.=C2=A0 If we&#39;re building a fallback, i=
t seems like these people shouldn&#39;t be excluded, since that means devel=
opers will have to build and maintain a 3rd fallback that&#39;s not WebTran=
sport based.=C2=A0 If the dream is to be able to develop apps for WebTransp=
ort and not have a non-WebTransport fallback, I don&#39;t see how we accomp=
lish that without an option over HTTP/1?</div><div><br></div><div>Possibly =
this is a forcing function to convince customers to upgrade their middlebox=
es to ones which support HTTP/2 and HTTP/3, but if so I&#39;d like to call =
that out explicitly.</div><div><br></div><div>Or maybe the physical office =
with its corp middleboxes is dead and my concern is an unfounded anachronis=
m?</div><div><br></div><div>Ian</div><div><br><div class=3D"gmail_quote"><d=
iv dir=3D"ltr" class=3D"gmail_attr">On Tue, May 25, 2021 at 5:36 PM Alan Fr=
indell &lt;afrind=3D<a href=3D"mailto:40fb.com@dmarc.ietf.org" target=3D"_b=
lank">40fb.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=3D"=
gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(20=
4,204,204);padding-left:1ex">I think the fundamental discussion here is abo=
ut whether we want to modify/extend the HTTP/2 state machine to handle WebT=
ransport streams, or whether we should build an entirely separate stream st=
ate machine for them, and leave the H2 machinery untouched.<br></blockquote=
></div></div></div></blockquote></div></blockquote></div></div></blockquote=
></div></blockquote></div></div></blockquote><div dir=3D"ltr"><br></div>Thi=
s sounds about right to me, if we think that the mapping of H3 onto H2 is n=
ot how we=E2=80=99d want it to be, we=E2=80=99ve evolved it significantly s=
o far, and I don=E2=80=99t see any reason why we can=E2=80=99t continue to =
do so.<br><blockquote type=3D"cite"><div dir=3D"ltr"><div class=3D"gmail_qu=
ote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bo=
rder-left:1px solid rgb(204,204,204);padding-left:1ex"><div class=3D"gmail_=
quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;=
border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><=
div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin=
:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"=
><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"marg=
in:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1e=
x"><div dir=3D"ltr"><div><div class=3D"gmail_quote"><blockquote class=3D"gm=
ail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,=
204,204);padding-left:1ex">
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br></blockquote>=
</div></div></div></blockquote></div></blockquote></div></div></blockquote>=
</div></blockquote></div></div></blockquote><div dir=3D"ltr"><br></div>I=E2=
=80=99d agree, I support adoption (as an author), and with that I=E2=80=99d=
 very much like to sit down and make sure we=E2=80=99re covering Martin=E2=
=80=99s feedback here =E2=80=94 I know parts of it were also raised at 110 =
prior to the interim and we haven=E2=80=99t all synced up to make sure it=
=E2=80=99s sorted out. That might result in minor changes to the document, =
it might result in major changes to the document =E2=80=94 either is fine f=
rom my perspective.</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">Thanks=
,</div><div dir=3D"ltr">Eric</div><div dir=3D"ltr"><br></div><div dir=3D"lt=
r"><br><blockquote type=3D"cite"><div dir=3D"ltr"><div class=3D"gmail_quote=
"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div class=3D"gmail_quo=
te"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bor=
der-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div=
 class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:0p=
x 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><d=
iv class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:=
0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">=
<div dir=3D"ltr"><div><div class=3D"gmail_quote"><blockquote class=3D"gmail=
_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204=
,204);padding-left:1ex">
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div>
</blockquote></div></div>
</blockquote></div>
</blockquote></div>
<span>-- </span><br><span>Webtransport mailing list</span><br><span><a href=
=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@ietf.org</=
a></span><br><span><a href=3D"https://www.ietf.org/mailman/listinfo/webtran=
sport" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a></span><br></div></blockquote></div></div></div></blockquote></div></di=
v>

--00000000000091f34705c3f107db--


From nobody Tue Jun  8 17:09:24 2021
Return-Path: <dschinazi.ietf@gmail.com>
X-Original-To: webtransport@ietfa.amsl.com
Delivered-To: webtransport@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 8CB2B3A11B5 for <webtransport@ietfa.amsl.com>; Tue,  8 Jun 2021 17:09:21 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -2.097
X-Spam-Level: 
X-Spam-Status: No, score=-2.097 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, FREEMAIL_FROM=0.001, HTML_MESSAGE=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Authentication-Results: ietfa.amsl.com (amavisd-new); dkim=pass (2048-bit key) header.d=gmail.com
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id dOeNCNT-N8Zt for <webtransport@ietfa.amsl.com>; Tue,  8 Jun 2021 17:09:16 -0700 (PDT)
Received: from mail-pj1-x1032.google.com (mail-pj1-x1032.google.com [IPv6:2607:f8b0:4864:20::1032]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id 8BC7E3A11AF for <webtransport@ietf.org>; Tue,  8 Jun 2021 17:09:16 -0700 (PDT)
Received: by mail-pj1-x1032.google.com with SMTP id mp5-20020a17090b1905b029016dd057935fso325584pjb.5 for <webtransport@ietf.org>; Tue, 08 Jun 2021 17:09:16 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025;  h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=GsMWN8FRQmWbXDZxHo5DBK6RW5XgeQRIxSzuYEwH2hE=; b=mGFSPezu5wIYXK5/FTjd+E8XLz5Zqk0QxkrIOSYpsabC2m8uo3jOb+NdfiCDG/QVxv +2GfEB1Qcieo3vrxo2chLMKGi48iBxl9n3UoeN7+Xni1O6HqTikr5XrqgZFqT6vxqmYa HW6VnF0dnoUnX+dOGTPklBuGISyFVIhFY5Fe3OFiumY8cdRIm5l3xHeWEJUnR84t6zMD H361ddT4m4BYjXwGTBUqV7chUgi9gcrpV5WzrehEwefv7xCBbUkZUtzegAYiMHuTYSFy JvoKqh+hFhRETBrrC0BUpTsBuuHgATGOy5mrbJlh3iPfjMtLx7gCR1jg949OtiBCU3Or BSQg==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=GsMWN8FRQmWbXDZxHo5DBK6RW5XgeQRIxSzuYEwH2hE=; b=OIck1xqwJo4x9ixyre0+WYNtOAuuuXt92Bs9zPP8Lo3ipVUNEiHQBY4IdfelQNjdC1 pfn70G952VT2knD23tAa4uYqeJq1O+qix/XMrKSbVIJIh4QbB9WVJEHyfYNmFO8MCFN9 dh0iZEO2i7osHx69IvRESfj7SfclbVkDKWra4Lv69193Orppn1B6blqvVGgQmwt1hiOP /u5xEvwJIBpJ/ULIlHHuqgXlHKWtXdGZ3qGE4oNLxUoD3O5q5NSuSg9Bggio2TtqcP/L RmZy/Xf1y2b1a1OSI6VWugqHFDA9gIvX7BFBbUzoA84pf0jdSDw+QYSJlUQCtW2YWzKg FcwQ==
X-Gm-Message-State: AOAM530Orym3lb/b706QBKKZp4ShAdf6+jyrIm/0iuMrcNXxkEB9jtM+ WBUninRAZgx6SUY7cGsVD0Wr3b2NXP6o9vmLAlU=
X-Google-Smtp-Source: ABdhPJyROGaLGiRvn19r0o+SsptZLofprKbEQ1KoFDTrEdHLmDx4bvol1nJqd3BErmWIMvTDnYbrSEOzfB877OvbUBs=
X-Received: by 2002:a17:90a:de07:: with SMTP id m7mr7583209pjv.100.1623197354228;  Tue, 08 Jun 2021 17:09:14 -0700 (PDT)
MIME-Version: 1.0
References: <7D4F22BF-6910-4E45-BC27-F395461B603C@apple.com> <CAKcm_gOywB_P9pRM3fUAdzJ+uh9+Snrwyz-0_zP+jXFhWNmrEw@mail.gmail.com>
In-Reply-To: <CAKcm_gOywB_P9pRM3fUAdzJ+uh9+Snrwyz-0_zP+jXFhWNmrEw@mail.gmail.com>
From: David Schinazi <dschinazi.ietf@gmail.com>
Date: Tue, 8 Jun 2021 17:09:03 -0700
Message-ID: <CAPDSy+7zbqJnw6PxBK+5X-SF7Dyd8Y0ikzz1GsKyFD2gh6fg-g@mail.gmail.com>
To: Ian Swett <ianswett@google.com>
Cc: Eric Kinnear <ekinnear@apple.com>, Victor Vasiliev <vasilvv@google.com>,  WebTransport <webtransport@ietf.org>, Alan Frindell <afrind@fb.com>,  Martin Thomson <mt@lowentropy.net>
Content-Type: multipart/alternative; boundary="000000000000b2fe4005c44a156b"
Archived-At: <https://mailarchive.ietf.org/arch/msg/webtransport/OUPOUY1knWCAWeEebtij5SorRQk>
Subject: Re: [Webtransport] Confirming consensus calls from today's WebTransport interim
X-BeenThere: webtransport@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: <webtransport.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/webtransport>, <mailto:webtransport-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/webtransport/>
List-Post: <mailto:webtransport@ietf.org>
List-Help: <mailto:webtransport-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/webtransport>, <mailto:webtransport-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 09 Jun 2021 00:09:22 -0000

--000000000000b2fe4005c44a156b
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hello everyone,

This consensus call is now concluded. The results are below:

Question 1: Do you believe that the WebTransport WG should start work
on a version of WebTransport that runs over TCP?
    1A: Yes, we should build this
    1B: No, we should not build this
The consensus was 1A.

Question 2: Do you believe that the version of WebTransport over TCP
should be over HTTP/2?
    2A: Yes, use HTTP/2
    2B: No, build a custom protocol over TLS/TCP
The consensus was 2A. The chairs note that this working group could
potentially also build a different version of WebTransport over TCP at
a later date if the need arises, but that would become the topic of a
separate discussion and consensus call.

Question 3: Do you believe that the WebTransport WG should adopt
draft-kinnear-webtransport-http2?
    3A: Yes, adopt
    3B: No, do not adopt
The consensus was 3A. The chairs note that there are some concerns
about how best to handle the differences between the HTTP/2 and
HTTP/3 stream state machines, however participants confirmed that
they were confident we could address those as a working group. The
document is adopted with the understanding that this topic will be
discussed very early on.

The chairs will work with the editors of that document to migrate it
to the working group GitHub organization, and will send details to
the list once that transition is complete.

Thanks,
David, for the chairs

On Fri, Jun 4, 2021 at 6:54 AM Ian Swett <ianswett@google.com> wrote:

> I've always thought QUIC over HTTP(or TCP) would be cool, but this is the
> first time I've heard a compelling use case.  Definitely worth considerin=
g.
>
> On Thu, Jun 3, 2021 at 2:17 PM Eric Kinnear <ekinnear=3D
> 40apple.com@dmarc.ietf.org> wrote:
>
>> =EF=BB=BFI think the question of HTTP/1.1 is an interesting one: we=E2=
=80=99ve so far
>> been operating with the view that falling back from QUIC to TLS/TCP migh=
t
>> enable significantly more folks to participate, but that the delta betwe=
en
>> HTTP/1.1 and HTTP/2 would not be as significant.
>>
>> I don=E2=80=99t have massively strong opinions in terms of how we go abo=
ut
>> addressing that, beyond the earlier concerns about additional effort
>> required to cover HTTP/1.1 vs. expected gain, but if the expected gain g=
oes
>> up, then perhaps that changes things sufficiently.
>>
>> It seems that some of that thinking has come from the observations made
>> about _who_ is in control of the versioning; if your network operator do=
es
>> not allow QUIC then that=E2=80=99s less under your control than whether =
HTTP/2 is
>> supported (although, as Ian mentioned, there=E2=80=99s a population with
>> middleboxes that force you down to HTTP/1.1, which might or might not
>> support CONNECT anyways).
>>
>> Would it be possible to gather some data about the relative size of thes=
e
>> different populations? I=E2=80=99m not sure there=E2=80=99s anyone in a =
great position to
>> do that, at least for CONNECT specifically if it=E2=80=99s not already
>> possible/in-use, but perhaps data about other H1 vs. H2 vs. H3 usage wou=
ld
>> be interesting.
>>
>
> Google has server-side data of "For people in an HTTP/3 enabled group,
> what fraction of people end up using HTTP/3, HTTP/2, and HTTP/1" and we c=
an
> breakdown by OS/etc.  I'm happy to share but I'd need approvals, so if I
> know exactly what data people would like to see, I can ensure it's all
> approved at once(ie: a single slide deck).
>
> This seems like a fun topic for maprg, but I'm not sure it's worth waitin=
g
> that long?
>
> It would be ideal if different people could share their data, since any
> single set of users may not be completely representative(ie: one could
> block a set of domains entirely, so none of the protocols worked).
>
>
>>
>> We know that, at least on iOS, HTTP/2 is in use in a higher percentage
>> for web browsing use cases than for other application use cases.
>>
>> On Jun 3, 2021, at 12:41 PM, Ian Swett <ianswett=3D
>> 40google.com@dmarc.ietf.org> wrote:
>>
>> =EF=BB=BF
>> Thanks David,
>>
>> I was thinking of corporate networks with MITMs that inspect HTTP 1.1 in
>> a limited way, but it sounds like those may not be very common, so there=
's
>> no way to make WebTransport work anyway.
>>
>> If so, making WebTransport HTTP/2 specific is fine with me.
>>
>> I still think apps that want to work close to 100% of the time won't be
>> able to depend upon WebTransport always working, but maybe that's just t=
he
>> state of the internet.
>>
>> Ian
>>
>> On Thu, Jun 3, 2021 at 12:28 PM David Schinazi <dschinazi.ietf@gmail.com=
>
>> wrote:
>>
>>> Can you provide an example of a network that would benefit from this
>>> scope increase? I've yet to see a network that MITMs TLS but allows all
>>> HTTP/1.1 through unmodified. The ones that MITM and downgrade to h1
>>> generally do it because they then inspect and/or modify the request and
>>> response, so getting something with WebTransport semantics to work ther=
e at
>>> all isn't guaranteed to be possible. Perhaps a good step forward is for=
 us
>>> to build a version of WebTransport over h2 and then look at deployment =
data
>>> and feedback to decide if another version is needed.
>>>
>>> David
>>>
>>> On Thu, Jun 3, 2021 at 8:56 AM Ian Swett <ianswett@google.com> wrote:
>>>
>>>>
>>>>
>>>> On Thu, Jun 3, 2021 at 11:38 AM David Schinazi <
>>>> dschinazi.ietf@gmail.com> wrote:
>>>>
>>>>> Hi Ian,
>>>>>
>>>>> Can you elaborate on what use-case would be solved by having
>>>>> WebTransport support HTTP/1 ?
>>>>> - We know there are networks that block QUIC and/or UDP, those will b=
e
>>>>> well served by HTTP/2.
>>>>> - We know there are networks that MITM TLS then force a transparent
>>>>> HTTP/1 proxy, those won't allow CONNECT nor any other form of WebTran=
sport
>>>>> anyway.
>>>>> I'm failing to see a use-case where WebTransport over HTTP/1 would be
>>>>> useful. Can you provide an example?
>>>>>
>>>>
>>>> I think there's evidence that those two sets of networks are very
>>>> similar today, and over time I expect the set of networks that block Q=
UIC
>>>> but don't block HTTP/2 to be tiny(ie: <1%).
>>>>
>>>> But I could be wrong.
>>>>
>>>>>
>>>>> Thanks,
>>>>> David
>>>>>
>>>>> On Thu, Jun 3, 2021 at 6:57 AM Ian Swett <ianswett=3D
>>>>> 40google.com@dmarc.ietf.org> wrote:
>>>>>
>>>>>> I definitely agree with 1a, with the caveat that the TCP version is
>>>>>> lower priority than the HTTP/3 version.
>>>>>>
>>>>>> I'm slightly less concerned about the stream state machine than
>>>>>> Martin, but I think it's definitely a real concern and I think it ma=
kes
>>>>>> sense to have a solution fairly well fleshed out before adopting the
>>>>>> direction and document.
>>>>>>
>>>>>> I'm slightly more concerned about creating a TCP fallback that
>>>>>> doesn't work over HTTP/1.  If we're building a fallback, it seems li=
ke
>>>>>> these people shouldn't be excluded, since that means developers will=
 have
>>>>>> to build and maintain a 3rd fallback that's not WebTransport based. =
 If the
>>>>>> dream is to be able to develop apps for WebTransport and not have a
>>>>>> non-WebTransport fallback, I don't see how we accomplish that withou=
t an
>>>>>> option over HTTP/1?
>>>>>>
>>>>>> Possibly this is a forcing function to convince customers to upgrade
>>>>>> their middleboxes to ones which support HTTP/2 and HTTP/3, but if so=
 I'd
>>>>>> like to call that out explicitly.
>>>>>>
>>>>>> Or maybe the physical office with its corp middleboxes is dead and m=
y
>>>>>> concern is an unfounded anachronism?
>>>>>>
>>>>>> Ian
>>>>>>
>>>>>> On Tue, May 25, 2021 at 5:36 PM Alan Frindell <afrind=3D
>>>>>> 40fb.com@dmarc.ietf.org> wrote:
>>>>>>
>>>>>>> I think the fundamental discussion here is about whether we want to
>>>>>>> modify/extend the HTTP/2 state machine to handle WebTransport strea=
ms, or
>>>>>>> whether we should build an entirely separate stream state machine f=
or them,
>>>>>>> and leave the H2 machinery untouched.
>>>>>>>
>>>>>>
>> This sounds about right to me, if we think that the mapping of H3 onto H=
2
>> is not how we=E2=80=99d want it to be, we=E2=80=99ve evolved it signific=
antly so far, and I
>> don=E2=80=99t see any reason why we can=E2=80=99t continue to do so.
>>
>>
>>>>>>> Martin may be right that there be dragons down the path of modifyin=
g
>>>>>>> the H2 state machine.  The alternative strikes me as an awful lot o=
f
>>>>>>> duplication -- it's close to implementing H2 again over a single HT=
TP
>>>>>>> stream, minus header processing.  That said, I think either approac=
h can be
>>>>>>> made to work, we just need to choose the tradeoffs we want.
>>>>>>>
>>>>>>> I am remiss in not addressing the feedback regarding unidirectional
>>>>>>> stream resets.  I think that it got lost during the bizarre hours o=
f the
>>>>>>> last IETF.  I filed an issue in the draft repo to track it (
>>>>>>> https://github.com/ekinnear/draft-webtransport-http2/issues/17).  I
>>>>>>> think the right solution is to keep the unidirectional reset functi=
onality,
>>>>>>> and implement it in the H2 draft using either new frames or message=
s,
>>>>>>> depending on the direction we take regarding the state machine.
>>>>>>>
>>>>>>> During the interim, we expressed our flexibility to adapt the desig=
n
>>>>>>> here as the H3 design evolves (eg: stream limits), new tools become
>>>>>>> available (eg: CAPSULE) and as we gather implementation experience.
>>>>>>>
>>>>>>> While I think it is important we address this design decision as
>>>>>>> soon as possible, I don't know that it needs to block adoption.
>>>>>>>
>>>>>>
>> I=E2=80=99d agree, I support adoption (as an author), and with that I=E2=
=80=99d very much
>> like to sit down and make sure we=E2=80=99re covering Martin=E2=80=99s f=
eedback here =E2=80=94 I
>> know parts of it were also raised at 110 prior to the interim and we
>> haven=E2=80=99t all synced up to make sure it=E2=80=99s sorted out. That=
 might result in
>> minor changes to the document, it might result in major changes to the
>> document =E2=80=94 either is fine from my perspective.
>>
>> Thanks,
>> Eric
>>
>>
>>
>>>>>>> Thanks
>>>>>>>
>>>>>>> -Alan
>>>>>>>
>>>>>>> =EF=BB=BFOn 5/23/21, 6:48 PM, "Webtransport on behalf of Martin Tho=
mson" <
>>>>>>> webtransport-bounces@ietf.org on behalf of mt@lowentropy.net> wrote=
:
>>>>>>>
>>>>>>>     On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:
>>>>>>>     > On Thu, May 20, 2021 at 9:37 PM Martin Thomson <
>>>>>>> mt@lowentropy.net> wrote:
>>>>>>>     > > I'm opposed to adoption of this document.  I don't believe
>>>>>>> that it can be made so due to some fundamental differences in how H=
TTP/2
>>>>>>> and QUIC stream states.
>>>>>>>     > >
>>>>>>>     > > If this problem was acknowledged, then I have every
>>>>>>> confidence that the people listed as authors are able to find a sol=
ution.
>>>>>>> But generally adoption is about the document more than the people.
>>>>>>> Adoption means agreeing that the document contains approximately th=
e right
>>>>>>> shape for a solution.  I can't agree with that in this case.
>>>>>>>     >
>>>>>>>     > Hi Martin,
>>>>>>>     >
>>>>>>>     > Could you elaborate on the specific differences between HTTP/=
2
>>>>>>> and QUIC
>>>>>>>     > stream state machines that you are concerned about?  The only
>>>>>>> one I am
>>>>>>>     > aware of is unidirectional versus bidirectional reset, and
>>>>>>> last time we
>>>>>>>     > discussed this, the easiest solution was to make HTTP/3 behav=
e
>>>>>>> like
>>>>>>>     > HTTP/2, which would mean there's nothing HTTP/2 draft would d=
o
>>>>>>> to
>>>>>>>     > address that.
>>>>>>>
>>>>>>>     That's the one I'm most concerned about, but there is also the
>>>>>>> whole mess around inventing semantics for unidirectional streams th=
at needs
>>>>>>> to be worked into the state machines.
>>>>>>>
>>>>>>>     For reset, sure, you could make the HTTP/3 draft worse to deal
>>>>>>> with the problem.  Why would you want that?  QUIC has a much better=
 set of
>>>>>>> semantics here than HTTP/2.
>>>>>>>
>>>>>>>     There is also the stream state management disconnect for
>>>>>>> server-initiated streams, which currently is only used for server p=
ush.
>>>>>>> The draft currently doesn't say anywhere near as much as it would n=
eed to
>>>>>>> in order to specify that part.
>>>>>>>
>>>>>>>     And managing allocation of stream limits is a little scant on
>>>>>>> detail in the same way.  I thought we wanted independent limits, wh=
ich
>>>>>>> means defining new accounting systems.  (Yes, this is a criticism o=
f the
>>>>>>> HTTP/3 draft also, but the additional effort required to use QUIC s=
treams
>>>>>>> is entirely justified in that case, because you are getting somethi=
ng in
>>>>>>> exchange.)
>>>>>>>
>>>>>>>     Why bother with all of that when you can just import QUIC frame
>>>>>>> semantics and then serialize them on the same stream?
>>>>>>>
>>>>>>>     Yes, this all boils down to a fairly central question: do we
>>>>>>> build the same thing for HTTP/2 and HTTP/3?  I'm not sure that we w=
ant
>>>>>>> that.  We need to do a whole bunch of stuff to use QUIC (BTW, Rober=
to gets
>>>>>>> to say "told you so" again here), but doing that same work for HTTP=
/2
>>>>>>> doesn't make sense unless the changes are minimal.  I'm suggesting =
that the
>>>>>>> changes aren't as minimal as the rough sketches of protocols we hav=
e might
>>>>>>> imply.
>>>>>>>
>>>>>>>     If you conclude that there are more functions needed, as I have
>>>>>>> done, then you might also conclude that building those same functio=
ns into
>>>>>>> two different protocols is not desirable, especially when there are=
 no
>>>>>>> benefits to doing so for one of those protocols.
>>>>>>>
>>>>>>>     --
>>>>>>>     Webtransport mailing list
>>>>>>>     Webtransport@ietf.org
>>>>>>>     https://www.ietf.org/mailman/listinfo/webtransport
>>>>>>>
>>>>>>> --
>>>>>>> Webtransport mailing list
>>>>>>> Webtransport@ietf.org
>>>>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>>>>
>>>>>> --
>>>>>> Webtransport mailing list
>>>>>> Webtransport@ietf.org
>>>>>> https://www.ietf.org/mailman/listinfo/webtransport
>>>>>>
>>>>> --
>> Webtransport mailing list
>> Webtransport@ietf.org
>> https://www.ietf.org/mailman/listinfo/webtransport
>>
>>

--000000000000b2fe4005c44a156b
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Hello everyone,<div><br></div><div>This consensus call is =
now concluded. The results are below:</div><div><br></div><div>Question 1: =
Do you believe that the WebTransport WG should start work<br>on a version o=
f WebTransport that runs over TCP?<br>=C2=A0 =C2=A0 1A: Yes, we should buil=
d this<br>=C2=A0 =C2=A0 1B: No, we should not build this<br>The consensus w=
as 1A.<br><br>Question 2: Do you believe that the version of WebTransport o=
ver TCP<br>should be over HTTP/2?<br>=C2=A0 =C2=A0 2A: Yes, use HTTP/2<br>=
=C2=A0 =C2=A0 2B: No, build a custom protocol over TLS/TCP<br>The consensus=
 was 2A. The chairs note that this working group could</div><div>potentiall=
y also build a different version of WebTransport over TCP at</div><div>a la=
ter date if the need=C2=A0arises, but that would become the topic of a</div=
><div>separate discussion and consensus call.</div><div><br>Question 3: Do =
you believe that the WebTransport WG should adopt<br>draft-kinnear-webtrans=
port-http2?<br>=C2=A0 =C2=A0 3A: Yes, adopt<br>=C2=A0 =C2=A0 3B: No, do not=
 adopt<br>The consensus was 3A. The chairs note that there are some concern=
s<br></div><div>about how best to handle the differences between the HTTP/2=
 and</div><div>HTTP/3 stream state machines, however participants confirmed=
 that</div><div>they were confident we could address those as a working gro=
up. The</div><div>document is adopted with the understanding that this topi=
c will be</div><div>discussed very early on.</div><div><br></div><div>The c=
hairs will work with the editors of that document to migrate it</div><div>t=
o the working group GitHub organization, and will send details to</div><div=
>the list once that transition is complete.</div><div><br></div><div>Thanks=
,</div><div>David, for the chairs</div></div><br><div class=3D"gmail_quote"=
><div dir=3D"ltr" class=3D"gmail_attr">On Fri, Jun 4, 2021 at 6:54 AM Ian S=
wett &lt;<a href=3D"mailto:ianswett@google.com">ianswett@google.com</a>&gt;=
 wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px =
0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=
=3D"ltr"><div>I&#39;ve always=C2=A0thought QUIC over HTTP(or TCP) would be =
cool, but this is the first time I&#39;ve heard a compelling use case.=C2=
=A0 Definitely worth considering.</div><br><div class=3D"gmail_quote"><div =
dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 2:17 PM Eric Kinnea=
r &lt;ekinnear=3D<a href=3D"mailto:40apple.com@dmarc.ietf.org" target=3D"_b=
lank">40apple.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote class=
=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rg=
b(204,204,204);padding-left:1ex"><div dir=3D"auto"><div dir=3D"ltr">=EF=BB=
=BFI think the question of HTTP/1.1 is an interesting one: we=E2=80=99ve so=
 far been operating with the view that falling back from QUIC to TLS/TCP mi=
ght enable significantly more folks to participate, but that the delta betw=
een HTTP/1.1 and HTTP/2 would not be as significant.<div dir=3D"ltr"><div d=
ir=3D"ltr"><br></div><div dir=3D"ltr">I don=E2=80=99t have massively strong=
 opinions in terms of how we go about addressing that, beyond the earlier c=
oncerns about additional effort required to cover HTTP/1.1 vs. expected gai=
n, but if the expected gain goes up, then perhaps that changes things suffi=
ciently.=C2=A0</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">It seems th=
at some of that thinking has come from the observations made about _who_ is=
 in control of the versioning; if your network operator does not allow QUIC=
 then that=E2=80=99s less under your control than whether HTTP/2 is support=
ed (although, as Ian mentioned, there=E2=80=99s a population with middlebox=
es that force you down to HTTP/1.1, which might or might not support CONNEC=
T anyways).=C2=A0</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">Would it=
 be possible to gather some data about the relative size of these different=
 populations? I=E2=80=99m not sure there=E2=80=99s anyone in a great positi=
on to do that, at least for CONNECT specifically if it=E2=80=99s not alread=
y possible/in-use, but perhaps data about other H1 vs. H2 vs. H3 usage woul=
d be interesting.</div></div></div></div></blockquote><div><br></div><div>G=
oogle has server-side data of &quot;For people in an HTTP/3 enabled group, =
what fraction of people end up using HTTP/3, HTTP/2, and HTTP/1&quot; and w=
e can breakdown by OS/etc.=C2=A0 I&#39;m happy to share but I&#39;d need ap=
provals, so if I know exactly what data people would like to see, I can ens=
ure it&#39;s all approved at once(ie: a single slide deck).</div><div><br><=
/div><div>This seems like a fun topic for maprg, but I&#39;m not sure it&#3=
9;s worth waiting that long?</div><div><br></div><div>It would be ideal if =
different people could share their data, since any single set of users may =
not be completely representative(ie: one could block a set of domains entir=
ely, so none of the protocols worked).</div><div>=C2=A0</div><blockquote cl=
ass=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid=
 rgb(204,204,204);padding-left:1ex"><div dir=3D"auto"><div dir=3D"ltr"><div=
 dir=3D"ltr"><div dir=3D"ltr"><br></div><div dir=3D"ltr">We know that, at l=
east on iOS, HTTP/2 is in use in a higher percentage for web browsing use c=
ases than for other application use cases.</div><div dir=3D"ltr"><br><block=
quote type=3D"cite">On Jun 3, 2021, at 12:41 PM, Ian Swett &lt;ianswett=3D<=
a href=3D"mailto:40google.com@dmarc.ietf.org" target=3D"_blank">40google.co=
m@dmarc.ietf.org</a>&gt; wrote:<br><br></blockquote></div><blockquote type=
=3D"cite"><div dir=3D"ltr">=EF=BB=BF<div dir=3D"ltr">Thanks David,<div><br>=
</div><div>I was thinking of corporate networks with MITMs that inspect HTT=
P 1.1 in a limited way, but it sounds like those may not be very common, so=
 there&#39;s no way to make WebTransport work anyway.</div><div><br></div><=
div>If so, making WebTransport HTTP/2 specific is fine with me.</div><div><=
br></div><div>I still think apps that want to work close to 100% of the tim=
e won&#39;t be able to depend upon WebTransport always working, but maybe t=
hat&#39;s just the state of the internet.</div><div><br></div><div>Ian</div=
></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr"=
>On Thu, Jun 3, 2021 at 12:28 PM David Schinazi &lt;<a href=3D"mailto:dschi=
nazi.ietf@gmail.com" target=3D"_blank">dschinazi.ietf@gmail.com</a>&gt; wro=
te:<br></div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px =
0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"=
ltr">Can you provide an example of a network that would benefit from this s=
cope increase? I&#39;ve yet to see a network that MITMs TLS but allows all =
HTTP/1.1 through unmodified. The ones that MITM and downgrade to h1 general=
ly=C2=A0do it because they then inspect and/or modify the request and respo=
nse, so getting something with WebTransport semantics to work there at all =
isn&#39;t guaranteed to be possible. Perhaps a good step forward is for us =
to build a version of WebTransport over h2 and then look at deployment data=
 and feedback to decide if another version is needed.<div><br></div><div>Da=
vid</div></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" class=3D"gma=
il_attr">On Thu, Jun 3, 2021 at 8:56 AM Ian Swett &lt;<a href=3D"mailto:ian=
swett@google.com" target=3D"_blank">ianswett@google.com</a>&gt; wrote:<br><=
/div><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bo=
rder-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><di=
v dir=3D"ltr"><br></div><br><div class=3D"gmail_quote"><div dir=3D"ltr" cla=
ss=3D"gmail_attr">On Thu, Jun 3, 2021 at 11:38 AM David Schinazi &lt;<a hre=
f=3D"mailto:dschinazi.ietf@gmail.com" target=3D"_blank">dschinazi.ietf@gmai=
l.com</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote" style=3D"ma=
rgin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:=
1ex"><div dir=3D"ltr">Hi Ian,<div><br></div><div>Can you elaborate on what =
use-case would be solved by having WebTransport support HTTP/1 ?</div><div>=
- We know there are networks that block QUIC and/or UDP, those will be well=
 served by HTTP/2.</div><div>- We know there are networks that MITM TLS the=
n force a transparent HTTP/1 proxy, those won&#39;t allow CONNECT nor any o=
ther form of WebTransport anyway.</div><div>I&#39;m failing to see a use-ca=
se where WebTransport over HTTP/1 would be useful. Can you provide an examp=
le?</div></div></blockquote><div><br></div><div>I think there&#39;s evidenc=
e that those two sets of networks are very similar today, and over time I e=
xpect the set of networks that block QUIC but don&#39;t block HTTP/2 to be =
tiny(ie: &lt;1%).</div><div><br></div><div>But I could be wrong.</div><bloc=
kquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:=
1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div><br></di=
v><div>Thanks,</div><div>David</div></div><br><div class=3D"gmail_quote"><d=
iv dir=3D"ltr" class=3D"gmail_attr">On Thu, Jun 3, 2021 at 6:57 AM Ian Swet=
t &lt;ianswett=3D<a href=3D"mailto:40google.com@dmarc.ietf.org" target=3D"_=
blank">40google.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquote clas=
s=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid r=
gb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div dir=3D"ltr"></div>I=
 definitely agree with 1a, with the caveat that the=C2=A0TCP version is low=
er priority than the HTTP/3 version.<div><br></div><div>I&#39;m slightly le=
ss concerned about the stream state machine than Martin, but I think it&#39=
;s definitely a real concern and I think it makes sense to have a solution =
fairly well fleshed out before adopting the direction and document.</div><d=
iv><br></div><div>I&#39;m slightly more concerned about creating=C2=A0a TCP=
 fallback that doesn&#39;t work over HTTP/1.=C2=A0 If we&#39;re building a =
fallback, it seems like these people shouldn&#39;t be excluded, since that =
means developers will have to build and maintain a 3rd fallback that&#39;s =
not WebTransport based.=C2=A0 If the dream is to be able to develop apps fo=
r WebTransport and not have a non-WebTransport fallback, I don&#39;t see ho=
w we accomplish that without an option over HTTP/1?</div><div><br></div><di=
v>Possibly this is a forcing function to convince customers to upgrade thei=
r middleboxes to ones which support HTTP/2 and HTTP/3, but if so I&#39;d li=
ke to call that out explicitly.</div><div><br></div><div>Or maybe the physi=
cal office with its corp middleboxes is dead and my concern is an unfounded=
 anachronism?</div><div><br></div><div>Ian</div><div><br><div class=3D"gmai=
l_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Tue, May 25, 2021 at 5:36=
 PM Alan Frindell &lt;afrind=3D<a href=3D"mailto:40fb.com@dmarc.ietf.org" t=
arget=3D"_blank">40fb.com@dmarc.ietf.org</a>&gt; wrote:<br></div><blockquot=
e class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px s=
olid rgb(204,204,204);padding-left:1ex">I think the fundamental discussion =
here is about whether we want to modify/extend the HTTP/2 state machine to =
handle WebTransport streams, or whether we should build an entirely separat=
e stream state machine for them, and leave the H2 machinery untouched.<br><=
/blockquote></div></div></div></blockquote></div></blockquote></div></div><=
/blockquote></div></blockquote></div></div></blockquote><div dir=3D"ltr"><b=
r></div>This sounds about right to me, if we think that the mapping of H3 o=
nto H2 is not how we=E2=80=99d want it to be, we=E2=80=99ve evolved it sign=
ificantly so far, and I don=E2=80=99t see any reason why we can=E2=80=99t c=
ontinue to do so.<br><blockquote type=3D"cite"><div dir=3D"ltr"><div class=
=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px =
0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div cla=
ss=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0p=
x 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div d=
ir=3D"ltr"><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" sty=
le=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);paddi=
ng-left:1ex"><div class=3D"gmail_quote"><blockquote class=3D"gmail_quote" s=
tyle=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);pad=
ding-left:1ex"><div dir=3D"ltr"><div><div class=3D"gmail_quote"><blockquote=
 class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px so=
lid rgb(204,204,204);padding-left:1ex">
<br>
Martin may be right that there be dragons down the path of modifying the H2=
 state machine.=C2=A0 The alternative strikes me as an awful lot of duplica=
tion -- it&#39;s close to implementing H2 again over a single HTTP stream, =
minus header processing.=C2=A0 That said, I think either approach can be ma=
de to work, we just need to choose the tradeoffs we want.=C2=A0 <br>
<br>
I am remiss in not addressing the feedback regarding unidirectional stream =
resets.=C2=A0 I think that it got lost during the bizarre hours of the last=
 IETF.=C2=A0 I filed an issue in the draft repo to track it (<a href=3D"htt=
ps://github.com/ekinnear/draft-webtransport-http2/issues/17" rel=3D"norefer=
rer" target=3D"_blank">https://github.com/ekinnear/draft-webtransport-http2=
/issues/17</a>).=C2=A0 I think the right solution is to keep the unidirecti=
onal reset functionality, and implement it in the H2 draft using either new=
 frames or messages, depending on the direction we take regarding the state=
 machine.<br>
<br>
During the interim, we expressed our flexibility to adapt the design here a=
s the H3 design evolves (eg: stream limits), new tools become available (eg=
: CAPSULE) and as we gather implementation experience.<br>
<br>
While I think it is important we address this design decision as soon as po=
ssible, I don&#39;t know that it needs to block adoption. <br></blockquote>=
</div></div></div></blockquote></div></blockquote></div></div></blockquote>=
</div></blockquote></div></div></blockquote><div dir=3D"ltr"><br></div>I=E2=
=80=99d agree, I support adoption (as an author), and with that I=E2=80=99d=
 very much like to sit down and make sure we=E2=80=99re covering Martin=E2=
=80=99s feedback here =E2=80=94 I know parts of it were also raised at 110 =
prior to the interim and we haven=E2=80=99t all synced up to make sure it=
=E2=80=99s sorted out. That might result in minor changes to the document, =
it might result in major changes to the document =E2=80=94 either is fine f=
rom my perspective.</div><div dir=3D"ltr"><br></div><div dir=3D"ltr">Thanks=
,</div><div dir=3D"ltr">Eric</div><div dir=3D"ltr"><br></div><div dir=3D"lt=
r"><br><blockquote type=3D"cite"><div dir=3D"ltr"><div class=3D"gmail_quote=
"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;borde=
r-left:1px solid rgb(204,204,204);padding-left:1ex"><div class=3D"gmail_quo=
te"><blockquote class=3D"gmail_quote" style=3D"margin:0px 0px 0px 0.8ex;bor=
der-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir=3D"ltr"><div=
 class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:0p=
x 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><d=
iv class=3D"gmail_quote"><blockquote class=3D"gmail_quote" style=3D"margin:=
0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">=
<div dir=3D"ltr"><div><div class=3D"gmail_quote"><blockquote class=3D"gmail=
_quote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204=
,204);padding-left:1ex">
<br>
Thanks<br>
<br>
-Alan<br>
<br>
=EF=BB=BFOn 5/23/21, 6:48 PM, &quot;Webtransport on behalf of Martin Thomso=
n&quot; &lt;<a href=3D"mailto:webtransport-bounces@ietf.org" target=3D"_bla=
nk">webtransport-bounces@ietf.org</a> on behalf of <a href=3D"mailto:mt@low=
entropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt; wrote:<br>
<br>
=C2=A0 =C2=A0 On Sat, May 22, 2021, at 02:06, Victor Vasiliev wrote:<br>
=C2=A0 =C2=A0 &gt; On Thu, May 20, 2021 at 9:37 PM Martin Thomson &lt;<a hr=
ef=3D"mailto:mt@lowentropy.net" target=3D"_blank">mt@lowentropy.net</a>&gt;=
 wrote:<br>
=C2=A0 =C2=A0 &gt; &gt; I&#39;m opposed to adoption of this document.=C2=A0=
 I don&#39;t believe that it can be made so due to some fundamental differe=
nces in how HTTP/2 and QUIC stream states.<br>
=C2=A0 =C2=A0 &gt; &gt; <br>
=C2=A0 =C2=A0 &gt; &gt; If this problem was acknowledged, then I have every=
 confidence that the people listed as authors are able to find a solution.=
=C2=A0 But generally adoption is about the document more than the people.=
=C2=A0 Adoption means agreeing that the document contains approximately the=
 right shape for a solution.=C2=A0 I can&#39;t agree with that in this case=
.<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Hi Martin,<br>
=C2=A0 =C2=A0 &gt; <br>
=C2=A0 =C2=A0 &gt; Could you elaborate on the specific differences between =
HTTP/2 and QUIC <br>
=C2=A0 =C2=A0 &gt; stream state machines that you are concerned about?=C2=
=A0 The only one I am <br>
=C2=A0 =C2=A0 &gt; aware of is unidirectional versus bidirectional reset, a=
nd last time we <br>
=C2=A0 =C2=A0 &gt; discussed this, the easiest solution was to make HTTP/3 =
behave like <br>
=C2=A0 =C2=A0 &gt; HTTP/2, which would mean there&#39;s nothing HTTP/2 draf=
t would do to <br>
=C2=A0 =C2=A0 &gt; address that.<br>
<br>
=C2=A0 =C2=A0 That&#39;s the one I&#39;m most concerned about, but there is=
 also the whole mess around inventing semantics for unidirectional streams =
that needs to be worked into the state machines.<br>
<br>
=C2=A0 =C2=A0 For reset, sure, you could make the HTTP/3 draft worse to dea=
l with the problem.=C2=A0 Why would you want that?=C2=A0 QUIC has a much be=
tter set of semantics here than HTTP/2.<br>
<br>
=C2=A0 =C2=A0 There is also the stream state management disconnect for serv=
er-initiated streams, which currently is only used for server push.=C2=A0 T=
he draft currently doesn&#39;t say anywhere near as much as it would need t=
o in order to specify that part.<br>
<br>
=C2=A0 =C2=A0 And managing allocation of stream limits is a little scant on=
 detail in the same way.=C2=A0 I thought we wanted independent limits, whic=
h means defining new accounting systems.=C2=A0 (Yes, this is a criticism of=
 the HTTP/3 draft also, but the additional effort required to use QUIC stre=
ams is entirely justified in that case, because you are getting something i=
n exchange.)<br>
<br>
=C2=A0 =C2=A0 Why bother with all of that when you can just import QUIC fra=
me semantics and then serialize them on the same stream?<br>
<br>
=C2=A0 =C2=A0 Yes, this all boils down to a fairly central question: do we =
build the same thing for HTTP/2 and HTTP/3?=C2=A0 I&#39;m not sure that we =
want that.=C2=A0 We need to do a whole bunch of stuff to use QUIC (BTW, Rob=
erto gets to say &quot;told you so&quot; again here), but doing that same w=
ork for HTTP/2 doesn&#39;t make sense unless the changes are minimal.=C2=A0=
 I&#39;m suggesting that the changes aren&#39;t as minimal as the rough ske=
tches of protocols we have might imply.<br>
<br>
=C2=A0 =C2=A0 If you conclude that there are more functions needed, as I ha=
ve done, then you might also conclude that building those same functions in=
to two different protocols is not desirable, especially when there are no b=
enefits to doing so for one of those protocols.<br>
<br>
=C2=A0 =C2=A0 -- <br>
=C2=A0 =C2=A0 Webtransport mailing list<br>
=C2=A0 =C2=A0 <a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">We=
btransport@ietf.org</a><br>
=C2=A0 =C2=A0 <a href=3D"https://www.ietf.org/mailman/listinfo/webtransport=
" rel=3D"noreferrer" target=3D"_blank">https://www.ietf.org/mailman/listinf=
o/webtransport</a> <br>
<br>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div></div></div>
-- <br>
Webtransport mailing list<br>
<a href=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@iet=
f.org</a><br>
<a href=3D"https://www.ietf.org/mailman/listinfo/webtransport" rel=3D"noref=
errer" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a><br>
</blockquote></div>
</blockquote></div></div>
</blockquote></div>
</blockquote></div>
<span>-- </span><br><span>Webtransport mailing list</span><br><span><a href=
=3D"mailto:Webtransport@ietf.org" target=3D"_blank">Webtransport@ietf.org</=
a></span><br><span><a href=3D"https://www.ietf.org/mailman/listinfo/webtran=
sport" target=3D"_blank">https://www.ietf.org/mailman/listinfo/webtransport=
</a></span><br></div></blockquote></div></div></div></blockquote></div></di=
v>
</blockquote></div>

--000000000000b2fe4005c44a156b--

