Re: rethinking caching

Jeffrey Mogul:
>I have added "Spoofing using Location headers (prevention thereof?)"
>to my list of issues for the caching subgroup, although this is not
>a commitment that we will actually solve the problem.

For the record, I feel that the spoofing using Location headers issue
is really a sub-problem of content negotiation, not of caching.  But
I'd be happy to deal with this problem in the caching subgroup instead
of the negotiation subgroup.  Two of the four members of the
negotiation subgroup are also in the caching subgroup, so there is
litte chance of these groups getting too much out of sync.

>I tend to agree with the view that this is not exactly a protocol
>design issue, but rather is a problem for people who are implementing
>shared web servers.

In my opinion, considering the impact of the protocol design
_is_ a protocol design issue, even if this consideration leads to a
review of shared web server security mechanisms.

We can only afford to introduce a web security problem and pass it to
the people who are implementing shared web servers after we have
established that these people can actually provide a solution to the
security problem.  As long as we do not have consensus that they can
solve the problem, we had better not put this security problem in the
HTTP protocol.

I have been thinking up a negotiation header structure that would not
allow any form of cache spoofing _and_ provide nice cache efficiency
(preemtive and reactive negotiation sharing cache slots).  The
downside of this structure would be the introduction of a new request
header with rather unusual semantics.  I'll try to post a description
in the near future.



Received on Tuesday, 19 December 1995 10:47:27 UTC