- From: Robert S. Thau <rst@ai.mit.edu>
- Date: Wed, 24 Apr 1996 16:00:40 -0400
- To: dmk@allegra.att.com, jg@w3.org
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
[ Warning --- half-baked idea follows ] Perhaps the spurious CRLFs following Netscape POST transactions would be easier to deal with if they were viewed as being appended to the beginning of the request following the POST rather than being appended to the end of the POST itself. That is, we could declare that in connections where HTTP/1.0 keep-alive back compatibility is desired, servers should allow a request-line to be *preceded* by an arbitrary amount of spurious white-space, including CRLF combinations, which they should simply ignore. I'm starting from the --- hopefully non-bogus --- theory that the CRLFs are in the stream no matter what, that we want the servers to ignore them, and the problem is coming up with a way of saying that without messing up the rest of the document. Unfortunately, the two suggestions floated so far do complicate the document a bit --- calling them "null requests" creates an exception to every rule elsewhere which "all requests" should follow, while considering them to be an addendum to the POST-request itself messes up the semantics of Content-length. The hope here, then, is that less of the document depends on the syntax of request-lines then on these other things, and so sweeping these CRLFs under that part of the rug, rather than another, results in a somewhat less unsightly bulge. However, there may very well be a dependance in the document that I've missed.... [ End half-baked idea. ] rst
Received on Wednesday, 24 April 1996 13:08:20 UTC