Re: HTTP response version, again

• From: Dave Kristol <dmk@research.bell-labs.com>
• Date: Mon, 23 Dec 96 09:36:04 EST
• To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
• Message-Id: <9612231436.AA04178@zp>

In the discussion so far the proponents of "send HTTP/1.1 in response
to an HTTP/1.0 request" emphasize the desirability of advertising the
server's capabilities.  But that shouldn't be necessary.  If a client
understands HTTP/1.1, it should send an HTTP/1.1 request (as Henrik
first noted).  An HTTP/1.0 server (with two known exceptions) will
respond with HTTP/1.0.  If a client sends an HTTP/1.0 request, then
sending an HTTP/1.1 response advertises something that's apparently of
no use to that client.

I agree with Dave Morris that we're trying to overload the meaning of
the protocol version in the status line.  If it's truly desirable to
advertise capability (and I'm not convinced), then I think the version
of the response and the capability version should be separated.  The
extra information would only be necessary when the two versions
differ.

Dave Kristol

Received on Monday, 23 December 1996 06:41:34 UTC