- From: Mike Cowlishaw <mfc@vnet.ibm.com>
- Date: Thu, 27 Apr 95 18:51:09 BST
- To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
- Cc: dmk@allegra.att.com
(Continuing after power-cut .. apologies for duplicated first sentences!) > Dave Kristol wrote: > I have to disagree: > 1) Only a few headers need opaque. We're not burdening all, or anywhere > close to all, of them. Of course, but once one header adds a special opaque field then everyone who 'owns' a header will add their own, too, citing the precedent. I'm looking long-term, here, and seeing (from bitter experience) the thin edge of a combinatorial wedge, where every header has every possible combination of attribute from every other header. > 2) It's a nuisance for the server to have to collate information from > two different headers. Not really. Most servers have to handle several headers already (Last-Modified-Since, Content-Type, Content-Length, etc.) Assuming there's a reasonable internal lookup-by-name routine, it's no more expensive to look up a separate header than to look up a field within a header. > In particular, if Opaque: has pieces that are > labeled for different other headers (if I understand what you're > proposing), then a server must, for example, parse both the Session-ID > and Opaque headers to figure out what's going on. It's much simpler to > keep all the information together and parse a single header. You're right for special-case parsing code, but that doesn't work, long term, or help much if you're asking a script-writer to handle new headers. > If you're assuming that there's a relationship between the value of > Opaque for Session-ID and the value for WWW-Authenticate, I think > you're wrong: I don't think they have any connection. Certainly > there's no obligatory connection. I think the value of Opaque -- due to its generality -- rather outweighs its value for either of those specific uses. At present, I'm adding hidden fields in fake forms in documents (which have no forms otherwise) simply to achieve the Opaque effect. I'd much rather do that at the HTTP level, where it belongs. Mike Cowlishaw
Received on Thursday, 27 April 1995 11:10:37 UTC