Mailing List Archive: Incorrect negotiation of encoding

Incorrect negotiation of encoding

Aug 9, 1995, 8:01 PM

Post #1 of 4 (242 views)

The current negotiation code is such that, if you request "somefile" and
only "somefile.gz" is available, then nothing is served.

This is because my browser doesn't send any Accept-Encoding (BTW, does
ANY browser send Accept-Encoding ?), and the negotiation code refuses to
serve an encoded version in that case.

But I think this is incorrect, as draft-ietf-http-v10-spec-01, chapter 9
(Content negotiation) says :

If the variant has no assigned Content-Encoding, or if no
Accept-Encoding field is present, the value assigned is "qe=1"

If you agree with me, then here's a patch :

Florent

*** mod_negotiation.c.orig Tue Aug 1 02:47:18 1995
--- mod_negotiation.c Wed Aug 9 19:45:54 1995
***************
*** 907,912 ****
--- 907,917 ----
var_rec *var_recs = (var_rec*)neg->avail_vars->elts;
int i;

+ /* If no Accept-Encoding is present, everything is acceptable */
+
+ if (!neg->accept_encodings->nelts)
+ return;
+
/* Lose any variant with an unacceptable content encoding */

for (i = 0; i < neg->avail_vars->nelts; ++i)

Re: Incorrect negotiation of encoding [ In reply to ]

rst at ai

Aug 9, 1995, 7:50 PM

Post #2 of 4 (244 views)

Permalink

Date: Wed, 9 Aug 95 20:01:44 +0200
From: guillaum@clipper.ens.fr (Florent Guillaume)
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

This is because my browser doesn't send any Accept-Encoding (BTW, does
ANY browser send Accept-Encoding ?), and the negotiation code refuses to
serve an encoded version in that case.

But I think this is incorrect, as draft-ietf-http-v10-spec-01, chapter 9
(Content negotiation) says :

If the variant has no assigned Content-Encoding, or if no
Accept-Encoding field is present, the value assigned is "qe=1"

If you agree with me, then here's a patch :

Look further --- Apache implements the content negotiation mechanism
specified in Roy's *previous* draft, which actually *mandates* the
objectionable behavior --- and this is far from the only thing which
changed. Just so you know...

rst

Re: Incorrect negotiation of encoding [ In reply to ]

fielding at beach

Aug 9, 1995, 8:52 PM

Post #3 of 4 (242 views)

Permalink

>Look further --- Apache implements the content negotiation mechanism
>specified in Roy's *previous* draft, which actually *mandates* the
>objectionable behavior --- and this is far from the only thing which
>changed. Just so you know...

And it will change again in two weeks -- this stuff is on the cutting
edge, so there's going to be quite a bit of bleeding in the near future.
It should be stable after that (he said knowingly... 8-0).

......Roy

Re: Incorrect negotiation of encoding [ In reply to ]

guillaum at clipper

Aug 10, 1995, 4:26 PM

Post #4 of 4 (238 views)

Permalink

> >[rst]
> >Look further --- Apache implements the content negotiation mechanism
> >specified in Roy's *previous* draft, which actually *mandates* the
> >objectionable behavior --- and this is far from the only thing which
> >changed. Just so you know...
>
>[Roy]
> And it will change again in two weeks -- this stuff is on the cutting
> edge, so there's going to be quite a bit of bleeding in the near future.
> It should be stable after that (he said knowingly... 8-0).

Okay, I'll wait.

But as content negotiation is one of the points where we don't need
compatiblity with NCSA, maybe we can allow ourselves to implement the
latest spec, no ?

Florent