- From: Rob Hartill <hartill@ooo.lanl.gov>
- Date: Tue, 15 Aug 95 14:26:25 MDT
- To: montulli@mozilla.com
- Cc: fielding@beach.w3.org, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Lou M writes, > If it is a violation of the spec then you really ought to make the spec > clearer on this point. All the major servers that I know of do a > greater than or equal comparison. The spec needs to be changed to > specifically say that 304 should only be returned when the If-modified-since > date matches the current modification date of the file. There's nothing to say that the If-modified-since is the date given to the client by the server it is subsequently sent to. Condsider a resource mirrored at A and B Server A can give you a Last-modified date for a URL which you can then use on server B. What you appear to be asking for (which might be useful as an alternative) is a If-something-different type header, e.g. a server sends Foo: 1234 client asks, If-Foo-Different: 1234 That might give everyone enough flexibility to increase the number of "304 Not Modified" messages that servers return. It's difficult for script writers to define Last-modified times for some output, but if they could negotiate on something simpler, with a straightforward comparison, then it'd encourage more script writers to write cache-friendly applications. I abuse the IMS system with lots of scripts by using it in the way Lou describes; returning a 304 only if the strings match (give or take a few formating differences). Writing a full IMS handler is beyond the capabilities of most CGI writers, and I'm just too lazy. rob -- http://nqcd.lanl.gov/~hartill/
Received on Tuesday, 15 August 1995 13:29:09 UTC