- From: Roy T. Fielding <fielding@kiwi.ICS.UCI.EDU>
- Date: Wed, 11 Jun 1997 09:27:16 -0700
- To: W.Sylwestrzak@icm.edu.pl
- Cc: http-wg@cuckoo.hpl.hp.com
>> > Unfortunately most of the servers practicing this today >> > try to perform a 'naive' content negotiation, which effectively >> > uses redirects to other urls. This is of course wrong, >> > because it unnecessarily expands the url addressing space, >> > thus making caching less effective. > >Drazen Kacar: > >> It isn't, because you'll suffer one redirection, but the target is >> cacheable. Compare it to serving different variants right away with >> ensuring that it won't be cacheable if the client is HTTP/1.0 compliant. > >I never claimed it's not cacheable. All I say is that this unnecessarily >multiplies the number of cached objects, which is bad. The number of cached objects is equal to the number of individually cachable responses retrieved through that cache, which is the exact same number whether they are one URL + Vary or multiple URLs. The only case where there is any savings is when the origin server is not sending Vary. In any case, it is rare for agent profiles to differ much behind a shared cache during the lifetime of a cached response. Using multiple URLs is advantageous because it gives the user agent more information, and thus more control over how it completes the retrieval, and also improves load-sharing and mirroring techniques and reduces the amount of information the client has to send to the server on every request. Using a single URL is advantageous because it only requires one round-trip, assuming the server guesses right. ....Roy
Received on Wednesday, 11 June 1997 10:30:39 UTC