W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2012

Re: HTTP/2 Expression of luke-warm interest: Varnish

From: Willy Tarreau <w@1wt.eu>
Date: Mon, 16 Jul 2012 22:46:50 +0200
To: Poul-Henning Kamp <phk@phk.freebsd.dk>
Cc: Roberto Peon <grmocg@gmail.com>, HTTP Working Group <ietf-http-wg@w3.org>, Mark Nottingham <mnot@mnot.net>, Mike Belshe <mike@belshe.com>
Message-ID: <20120716204650.GI27010@1wt.eu>
On Mon, Jul 16, 2012 at 08:14:09PM +0000, Poul-Henning Kamp wrote:
> In message <CAP+FsNdJciE+r6PiyBzh4tgpDj0Vg6+9UkuF_7c+9ZMjgzo7Yg@mail.gmail.com>
> , Roberto Peon writes:
> 
> >> What we should do now is throw the prototype away, and design a
> >> serialization of HTTP which is as if not more efficient than deflate,
> >> and which is less hostile to HTTP routers less prone to DoS
> >> exploitation, and takes less memory footprint (code&data) to implement.
> >
> >We're all here to try to do that, as far as I know.   :)
> 
> Well, that's not the vibe that transmits all the way over here, from
> here it sounds like: "adopt SPDY as the starting point, rearrange
> the deck-chairs a bit then call it HTTP/2.0"
> 
> But the proof is in the pudding:  The best way to shut me up is to do
> a damn good job.

Poul-Henning, we're *really* here to improve this, I mean *really*.

When we started with Amos, I said "let's not waste our time redesigning
flow control, SPDY already has it, let's address what we don't agree with
first". We worked hard to improve a few points. The result is that we don't
have a complete proposal but something to experiment with, a PoC (and a
lib almost ready in Amos' bag).

I don't care a dime *whose* proposal is used as a basis, leave this to the
press who only sees a fight between google and microsoft to own HTTP/2. Facts
are that SPDY has the most experience. Others brought different ideas. I like
Gabriel's idea of trying to build a common layer for WS and HTTP eventhough I
don't like much the first proposal. I like the experiments we have done
consecutive to our draft eventhough I don't like at all how it's proposed
right now. I liked a lot Roy's presentation of Waka and am terribly sad it's
still not in a draft.

This is about making progress through experimentation. I mean, even if in
the end we adopt 100% of SPDY after all other attempts have failed, then we
will know it was the best tradeoff. Right now I'm far from being convinced
of that and that's why I'm motivated in experimenting with other tracks.
And I'd say even the SPDY guys are open to this.

So we're here to make engineering work, to vigorously defend our ideas,
fears and expectations, but only on technical grounds.

It's perfectly expected that some early adopters report a big success.
We should just consider this with a grain of salt when they're dealing
with monster datacenters, but they represent a valid use case. Even some
early adopters had mixed opinions, that's expected too. Some developers
have fears that are expected. All what happens here is normal and it's
not because 3-4 very satisfied implementers support one protocol that
it suddenly the protocol of everyone. However it could become de facto
if the WG cannot make progress to standardize something !

We're not here to win anything, we're here to improve HTTP so that we
don't have to implement something which causes more issues than it solves,
because in the end, we will *all* have to implement HTTP/2.0, so we must
like it first. *This* is what I'm doing here and what I hope most of us
are doing here.

It's also why I'm afraid of the timing BTW, some of us difficultly have
more time to spend on this and it's a shame because some experiments are
needed.

I also remember that in private with a few of us you proposed a draft
that you never published. There were very valid concerns here. Why do
you not share it so that good ideas can be picked from it ? It would
help a lot more than attacking the proposal that already gets support
for already being deployed at some places.

Regards,
Willy
Received on Monday, 16 July 2012 20:47:23 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 16 July 2012 20:47:30 GMT