- From: Koen Holtman <koen@win.tue.nl>
- Date: Wed, 15 May 1996 10:27:56 +0200 (MET DST)
- To: Larry Masinter <masinter@parc.xerox.com>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Larry Masinter: [...] >some more steps here ... but ultimately leading to the conculsion >that: > >THEREFORE the protocol must support caching of multiple entities for >the same URL, in that a proxy may return different fresh entities for >the same URLs as long as the proxy determines that the request headers >of the subsequent requests match the appropriate request headers of >the original request that evoked the original entity. I agree. My report on the Paris stuff did not intend to contradict this requirement. It merely discussed a new way of talking about these multiple entities which we want to be cached for a single URI. Instead of a model in which these entities are bound directly to the generic resource, it outlined a model in which they are bound to variant resources which in turn are bound to the generic resource. Caching proxies would still be matching request headers: the only difference is that in this model, the proxy would match request headers to choose the appropriate variant resource, instead of choosing the appropriate entity. >I think the only difficulty comes when a resource might have several >(stale) entities associated with it in a given cache and the cache >recieves a new fresh entity, it isn't clear which of the old stale >entities might be discarded. Personally, I think this is a cache >optimization issue and not a protocol correctness issue; I can think >of several heuristics that a cache might employ to do a reasonable job >in such a case. We have been through this before: it is a efficiency issue, but most of us want the protocol to contain a mechanism which helps caches to be more efficient. This mechanism could either be a variant-ID or a Content-Location header. The `several heuristics' possible if such a mechanism is lacking are not considered to be good enough by most of us. >Larry Koen.
Received on Wednesday, 15 May 1996 01:27:58 UTC