- From: Anselm Baird_Smith <abaird@www43.inria.fr>
- Date: Tue, 3 Dec 1996 19:14:08 +0100 (MET)
- To: Steve Wingard <swingard@spyglass.com>
- Cc: Dave Kristol <dmk@research.bell-labs.com>, masinter@parc.xerox.com, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Steve Wingard writes: > At 12:05 PM 12/3/96 EST, Dave Kristol wrote: > >I don't recall whether the following issue was resolved on the mailing list: > > > >What protocol version number should an HTTP/1.1-compliant origin server > >send for an HTTP/1.0 request? > > > >There seemed to be two camps: > >1) Send HTTP/1.0 as the response to HTTP/1.0 requests (and HTTP/1.1 as the > > response to HTTP/1.1 requests). > > > > Pro: HTTP/1.0 clients may only understand HTTP/1.0 responses > > Con: a client would never be able to determine whether a server > > understands HTTP/1.1 > > > >2) Send HTTP/1.1 responses always. > > > > Pro: the server advertises its capability > > Con: because the response (headers) must be HTTP/1.0 > > compatible, the server is "lying" about the kind of > > response and may mislead or confuse the client. > > > >My preference is (1). > >Dave Kristol > > > > We've been working with implementation #2 (taking care not to use > any 1.1 mechanisms that would cause problems to a 1.0 client when > the request indicates HTTP/1.0), and have not encountered any > interoperability issues yet. I've noticed that www.apache.org is running > a preliminary version of Apache v1.2 that returns HTTP/1.1 in its > responses -- Robert, have you folks gotten any complaints from > any users? > > Has anybody else done any "experimentation"? Jigsaw uses (2) to; The proxy had some problems (some old NCSA servers), but nothing really ennoying. My idea was to send the highest minor number within the match of the major number of the client. If major number changes, I would probably have to use (1), though (another story, anyway, since Upgrade would probably be involved too). Anselm.
Received on Tuesday, 3 December 1996 10:31:46 UTC