Re: Problems with content negotiation (was: Re: Preemptive and reactive content negotiation)

The problem in content negotiation is that 

   * most documents are available only in one format
   * fully disclosing accepted types takes lots of protocol

So, what if client sends to server 'accept-signature' which is a hash
(MD5 or something smaller) of all headers that are relevant to content
negotiation.

Server either has seen / remembers this signature or not. If not, AND
if the document is actually available in multiple formats, the server
sends back a response asking for expansion of accept information.

Client sends back full accept headers, server remembers them (for a
while) and then uses the information in content negotiation.

The 'accept-signature' doesn't change rapidly; many like-configured
browsers will have the same signature; servers that get multiple hits
from the same or similar clients won't have to recompute/reinterpret
the signature; servers that don't need to do content negotiation
ignore it all, the headers are small most of the time.
The signature for embedded images can be different than the signature
for regular links.

This could even be combined with smaller indications of the 'main'
type.

Received on Sunday, 10 September 1995 03:26:54 UTC