- From: Paul Leach <paulle@microsoft.com>
- Date: Wed, 6 Sep 95 14:09:52 PDT
- To: mogul@pa.dec.com
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Jeff said: ] Frankly, it doesn't much matter to me if we use "Expires: yesterday" ] or "Cache-control: always-validate"; these are just syntactically ] different ways of expressing the same thing, IF (note the big IF) ] we agree on the basic rule that ] ] for an object with Expiration time T ] prior to time T, the cache is assumed to be valid ] after time T, validity MUST be checked with the server I'll agree 100% that it doesn't matter to me either, if we can add the following rider: IF "Cache-control: always-validate" is present, THEN "Expires:" is not allowed Otherwise, what do I do if both are present? That would seem to break your rule above, and necessitate changing it to: for an object with Expiration time T prior to time T, the cache is assumed to be valid, unless "Cache-control: always-validate" was present when it was fetched after time T, validity MUST be checked with the server, unless "Cache-control: always-validate" was present when it was fetched, in which case it must be refetched ] Arguments about what is "natural" or not are pretty pointless, since ] it appears that two people have diametrically opposite ideas of what ] is natural here. I agree. I should choose words more carefully. I was only claiming "naturalness" with respect to a particular sketch of how caches might work, where (I at least hoped) it was clear that the intended behavior fell out with very few changes. I make no claim that other people would find that sketch reasonable, or whether they would agree that that's the right metric of "natural". But at least people would know why I thought it was more natural. ] I would point out that the "Expires: yesterday" ] approach might be more likely to cause correct operation of existing ] proxies, but I don't have specific evidence to back this up. I guess if their general approach to managing their cache was close to the sketch, then it would be likely. Paul
Received on Wednesday, 6 September 1995 14:55:45 UTC