Re: ETags vs Variants, was: Revising RFC2616 - what's happening

sön 2006-11-05 klockan 17:34 -0800 skrev Roy T. Fielding:

> There is no issue here.  Vary works in many implementations and there
> has never been a single report of interoperability problems between
> clients and servers that have implemented Vary as specified.  It is an
> integral part of HTTP/1.x caching that cannot be deprecated.

While I agree with you in principle I am still a bit curious. Can you
name an implementation where Vary really works?

MSIE -> no-store

Firefox -> Stores at most one variant per URL. Broken if no ETag in the
response. In such case it's processed as kind of a no-cache and Sends
If-Modified-Since based on the old response, and the real error being
taking a 304 response to mean that the old response is valid for the new
request headers as well. ETag + If-None-Match prevents the bug from
surfacing if supported by the server.

Squid -> kind of attempts to obey Vary, but with limitations

Delegate -> no-store

Regards
Henrik