httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Alexei Kosut <>
Subject Re: HTTP/1.1 and colons and things
Date Thu, 20 Jun 1996 17:09:22 GMT
On Thu, 20 Jun 1996, Roy T. Fielding wrote:


> It is broken on CERN, though it will give the default choice
> (not always the right choice) instead of an error.  For testing,

Ah. I did not know this. Well, that's good, I s'pose. I guess the CERN
people were following an even earlier spec than the Apache people. Though,
in your example (for example), it's just as bad as an error, because not
many browsers that I know of correctly allow TIFFs as inline images.


>    HEAD /~fielding/hidden/m1 HTTP/1.0
>    Accept: image/jpeg:q=1
>    HTTP/1.0 404 Not found
>    Date: Thu, 20 Jun 1996 08:08:02 GMT
>    Server: Apache/1.0.5
>    Content-type: text/html
> YIKES -- that isn't even the right response code (400 or 406 is).

The comment in the code, btw, says:

 * TO DO --- error code 406.  Unfortunately, the specification for
 *           a 406 reply in the current draft standard is unworkable;
 *           we return 404 for these pending a workable spec. 

This comment goes all the way back to Apache 0.6.x (at least).
Unfortunately, it's thirteen months later and there *still* isn't a
workable 406 spec. (even Koen's new content negotiation draft, which has  
a workable 300 spec, doesn't cover 406).

At any rate, I guess whomever wrote it at the time figured that 404 would
be appropriate. Doesn't matter now, though. 400, 404, 406, they all mean
the same thing at the moment.


> because the first ":" does not affect the latter Accept value, but
>    HEAD /~fielding/hidden/m1 HTTP/1.0
>    Accept: image/jpeg:q=0.5, image/gif:q=0.8
>    HTTP/1.0 404 Not found
>    Date: Thu, 20 Jun 1996 08:10:48 GMT
>    Server: Apache/1.0.5
>    Content-type: text/html
> dies miserably.

Yes. This is the icky case. And, because current browsers (e.g. Netscape)
actually send workable Accept: headers for inline image requests
(they don't send q-values, but they do send the image types they
understand), a lot of people are actually using content negotiation in
Apache in this way.

> I don't consider (1) as an option any more -- looks like the HTTP/1.1
> draft must be updated to prevent extreme hell from breaking loose.

Unfortunately agreed. I don't see any good way around it.

-- Alexei Kosut <>            The Apache HTTP Server

View raw message