Re: h2 priority

On Tue, Sep 2, 2014 at 11:51 AM, James M Snell <jasnell@gmail.com> wrote:

>
> > The prototype protocol now carries over 1/3 of my HTTP transactions in
> > production - many others are waiting for a stable and open standard to
> work
> > with before joining the party. The only question left is really whether
> the
> > IETF is institutionally capable of providing that standard or another
> venue
> > will have to step up and fill that gap in this case. I certainly hope it
> can
> > - we'll find out soon.
> >
>
> So, in other words, "If you don't play how we want to play, we'll just
> take our ball somewhere else!" That hardly seems helpful or "open".
>

James, I do not appreciate you saying 'in other words' and using quotation
marks in attempt to paraphrase me. I said what I said. You are free to say
what you like. In any event, I strongly disagree with your characterization.

The open h2 process has resulted in innumerable changes from spdy. Many of
them are good - the total sum of them certainly did not require 2.5 years
to work out, but the process of being open may have.  However, the largest
change is probably the mistake to remove the mandatory TLS binding. But I'm
still here even though I deeply disagree with that. Indeed there have been
some big disagreements over the last couple of years and I don't recall
anyone walking away though I'm sure we all have certain things that would
make the end result not worth shipping.

As we move through the process the bar for making a change is raised -
we've talked about that a number of times.

To clarify my statements, I indicated that the world is moving beyond
HTTP/1 - the data shows that. We've also got an existence proof that the
folks involved in that transition have sought to participate in an IETF
forum to produce an open standard around that transition. The consensus of
the group was to base h2 on spdy making changes as necessary. That's where
we've been for 2.5 years. As I said, what's left to be seen is whether or
not an open standard actually comes out the other end of the process. We've
made it to LC - now we need the discipline to separate substantive from
nice-to-have.

The WG or IAB/IESG could of course choose to standardize something
different than h2-14, or even decide that the basic prioritized muxx'ed
properties of h2 on tcp aren't something that should be standardized no
matter how widely used. That judgment is an important part of their roles
as a standard is at least partially an endorsement. I don't mean to
pre-suppose its outcome. If no open standard is produced the action will
likely move elsewhere because the need will not have dissipated. In the big
picture that's OK, in the immediate picture its an insane waste of 2.5
years of energy. And no, I don't expect that will happen - we just need to
be focused down the home stretch.

either. However, here we have some very legitimate, non-editorial
>
>
all feedback is important. And here I speak generically - not about Roy's
emails (where I wish to comment on them specifically, I will comment on
them specifically as I have elsewhere in this thread :)) - Last Call isn't
the best time to re-open issues that have been tracked, discussed, and
resolved previously.