Re: draft-yevstifeyev-http-headers-not-recognized-01.txt

23.11.2010 12:37, Sylvain Hellegouarch wrote:
> On Tue, Nov 23, 2010 at 11:13 AM, Julian Reschke 
> <julian.reschke@gmx.de <mailto:julian.reschke@gmx.de>> wrote:
>
>     On 23.11.2010 10:16, Mykyta Yevstifeyev wrote:
>
>         ...
>
>         If a server sends 4xx code, it stops procesing the request.
>         Generally unrecognized headers are not critical - all
>         'vital' headers MUST be supported by servers. If a client receives
>         such a response, it SHOULD avoid sending requests
>         with mentioned headers while server MUST cntinue processing
>         the request.
>         ...
>
>
>     *Why* "SHOULD" it avoid sending those? What's the problem you're
>     trying to solve? Wasted bits? It would be helpful to understan
>
>
> Also I'm afraid we'd end up with a ping/pong game where clients would 
> likely dismiss the new status code anyway as servers ignore currently 
> headers they don't understand.
IMO it wouldn't be a great trouble for servers to add another one HEADER 
(not separate
status code!) with a list of headers it hasn't recognized. The aim of 
this header is to
inform the client about not processed headers. However there is no MUST 
criterion
for avoiding sending the 'not-recognizable' headers. If the client wants 
them to be sent,
it will sent, but it is only RECOMMENDED that these headers are not sent 
(siad in other
words).
Moreover, if a sever sends the corresponding headers, a client MAY try 
to change it
to acceptable one (for server) in case it is critical for it (client).
Finally, server SHOULD (and not MUST) send the 'Headers-Not-Recognized' 
Header.
If it is a big difficulty for them (but it's hard to imagine), they 
won't do this.
I don't understand what all this fuss is for?

Mykyta Yevstifeyev
>
> -- 
> - Sylvain
> http://www.defuze.org
> http://twitter.com/lawouach