Re: [hybi] Restarting IESG review on permessage-deflate

[Takeshi Yoshino <[e-mail]>]

On Fri, Oct 4, 2013 at 11:24 PM, Joakim Erdfelt <[e-mail]> wrote:

> There is also Eclipse Jetty 9.1 (currently in RC releases) support (server
> and client)
>
> We cannot support some of the spec, as written.
> Our experiences implementing it, and using chromium + pywebsocket for
> testing both the server or client implementation.
>
> The following configuration parameters are not supported, nor can/will be.
>
> s2c_no_context_takeover
> c2s_no_context_takeover
> s2c_max_window_bits
> c2s_max_window_bits
>
>
Sorry, why? s2c_no_context_takeover requests the server to reset context.
The server just need to throw away current deflater, create a new one and
use it.

If the client didn't send s2c_no_context_takeover, you can still do the
same. You don't have to carry over compression context.

The problem is that your server cannot decompress incoming data if the
client carries over compression context. So, you can just reply with
c2s_no_context_takeover. In
http://tools.ietf.org/html/draft-ietf-hybi-permessage-compression-12#section-8.1.1.2,
a client is required to implement c2s_no_context_takeover. I made this
mandatory since it's doable. So, you don't have to worry about clients
coming without c2s_no_context_takeover capability.


> This level of control of the deflate algorithm is just not possible in
> Java, sorry.
>
>
Again, please explain why you can't support no_context_takeover family
parameter.