Re: [hybi] [permessage-deflate] Renaming s2c_ and c2s_ prefix to server_ and client_

[Takeshi Yoshino <[e-mail]>]

On Tue, Oct 22, 2013 at 12:50 AM, Tobias Oberstein <
[e-mail]> wrote:

> AutobahnPython 0.6.4 has been adjusted and published:
>
> https://pypi.python.org/pypi/autobahn
> https://github.com/tavendo/AutobahnPython/blob/master/CHANGELOG.md
>
> Tested with canary 32.0.1677.0: no issues found .. I've tested a couple of
> negotiated parameter sets also.
>
> We will extend AutobahnTestsuite for _systematic_ testing of various
> parameter sets ..
>

Thank you!


> From my side, there is only 1 minor point left .. see below. I just want
> to raise it for completeness, but am happy if Draft-15 makes it into final
> RFC also.
>
> Sidenote: I've seen that Chrome will log extensions negotiated in Web
> inspector also. Nice nice. Thanks. It seems just hours away .. mine does
> not have it yet. Anyway: reckless pace! Awesome.
>

Yes. It has been unexpectedly dropped. We've fixed it recently.


>
> /Tobias
>
>
> Proposal: Make support for "no context takeover" optional, and
> consequently announce "client_no_context_takeover".
>
> That makes the whole parameter negotiation more "consistent" and allows
> implementors who do not want to implement "no context takeover" for
> whatever reasons (technically, it should be possible to do always, but ..).
>
>
As you know, we've given the role as a hint to client_no_context_takeover
parameter in a negotiation offer, that syntax is not available for
capability announcement now. So, we need to introduce a new syntax for hint
like:

client_no_context_takeover=likely

to free the syntax "client_no_context_takeover without a value" for
capability announcement. Or instead, we need to drop the hint feature.

Maybe Java users like Joakim would be happier with having MUST there than
SHOULD?


> Chrome announces:
>
> 1. Sec-WebSocket-Extensions:
> permessage-deflate; client_max_window_bits, x-webkit-deflate-frame
> and supports "client_no_context_takeover", but that is implicit, not
> explicit.
>
> Autobahn (client) currently (deliberately) announces:
>
> Sec-WebSocket-Extensions: permessage-deflate; client_no_context_takeover;
> client_max_window_bits
>
> hence _explicitly_ announces "client_no_context_takeover"
>
> ===
>
> Should above _not_ go into the final spec, AutobahnPython will then of
> course comply (and remove the announcement) when the RFC is issued.
>