Re: Priorities I-D for Thursday HTTPbis meeting (Fwd: New Version Notification for draft-kazuho-httpbis-priority-04.txt)

[Robin MARX <[e-mail]>]

Hello Martin,

 > I understand that reprioritization was identified as being important by
the design team, but I don't recall seeing ever any evidence to support
that view.  Can we split that piece out and maybe pursue it separately?

Could you expand on this statement a bit more?
In my view, reprioritization has two sides: reprioritization from the
client after sending an initial priority and server-side reprioritization,
overriding the client's initial priority.

The first side comes into play when doing things like video/media streaming
when switching qualities (see e.g.,
https://github.com/kazuho/draft-kazuho-httpbis-priority/issues/92) or for
the "switching tabs" use case that e.g., Ian keeps mentioning.

The second side was discussed at some length in our paper (
https://h3.edm.uhasselt.be/files/HTTP3_Prioritization_extended_3jul2019.pdf)
and also in a series of Cloudflare blogposts (e.g.,
https://blog.cloudflare.com/better-http-2-prioritization-for-a-faster-web/).
It is also imo mirrored in the intent of the priority hints proposal from
the W3C (https://github.com/WICG/priority-hints/blob/master/EXPLAINER.md).
One of the main reasons for simplifying the H2 scheme (next to
implementation complexity) is to provide an "absolute" priority indication
so that it's easier to reprioritize the responses on the server side.

Are you talking about both or one of these sides?
While I would agree we have less data for the 1st side, I do feel everyone
on the design team agrees this should be kept in.
Removing the PRIORITY_UPDATE frame would also lead to significant push back
from the Chrome team, as I understood from Ian (though I don't want to put
words in his mouth), and would make the proxy-to-origin case difficult
again.

With best regards,
Robin


On Thu, 21 Nov 2019 at 10:00, Martin Thomson <[e-mail]> wrote:

> On Thu, Nov 21, 2019, at 09:53, Roy T. Fielding wrote:
> > Even with these minor issues, I think this draft is a great improvement
> > and is ready for adoption by the WG.
>
> Oh, I forgot this.  I agree with Roy, we should definitely adopt this
> now.  The issues I raise are minor enough that we can work them out in the
> working group.
>
>

-- 

Robin Marx
PhD researcher - web performance
Expertise centre for Digital Media

T [phone] - GSM [phone]

www.uhasselt.be
Universiteit Hasselt - Campus Diepenbeek
Agoralaan Gebouw D - B-3590 Diepenbeek
Kantoor EDM-2.05