Re: SPDY Header Frames

On Fri, Jul 13, 2012 at 11:01:15PM +0000, Poul-Henning Kamp wrote:
> In message <20120713225104.GK16256@1wt.eu>, Willy Tarreau writes:
> 
> >Whatever will be retained as a basis for HTTP/2.0, this exercise is
> >useful and may incite other users to provide very valuable feedback.
> 
> I think it is premature, because it obscures and prevents the much
> needed high-level design of HTTP/2.0.
> 
> And that is exactly why I think the current approach and timeline
> is a road to nowhere fast.

On a personal taste, I find it fast too. 4 months to provide proposals
to replace the 15-year old HTTP/1, and 4 others to review them is short
in my opinion. Roy did not even have the time to publish the Waka spec
which could have brought a lot of fuel to the discussion !

> The fact that we just saw Google say they would get behind any
> improvement, as long as it is SPDY pretty much dooms the HTTP/2.0
> effort right there and then:  All that's on the table is minor
> rearrangements of the deckchairs, there is no opening for
> changing the course.

If you're talking about Roberto's mail, I read quite the opposite in
fact. It was said that Google was open to have anything provided that
the concepts raised in SPDY were preserved. By this I understand that
they want something that offers similar end user experience. I find
it much more open than how you read it.

I have talked long hours with the SPDY team at IETF83. They have running
code, users and data. Some of us (including me) don't like the way they
addressed certain things, especially in the context of HTTP/2. They
clearly said they were open to changes. What else do you want ? That's
why we're working on selected points that we think need to be addressed
instead of reinventing the wheel from scratch.

With more time, it would probably make sense to discuss every single point
here on the WG as was done for WebSocket with nothing stabilizing for 1
year, because a group-designed protocol should be much better than the one
designed by a small team to address a specific issue. But I think it still
makes sense to use the proposals as a basis for new work, eventhough some
concepts are missing from all of them (eg: user session).

Regards,
Willy

Received on Saturday, 14 July 2012 05:30:00 UTC