- From: Andrew McRae <mcrae@elmer.harvard.edu>
- Date: Sat, 7 Oct 1995 16:08:40 -0400 (EDT)
- To: Paul Hoffman <ietf-lists@proper.com>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Hi, all.
On Sat, 7 Oct 1995, Paul Hoffman wrote:
> This is true for present software: the server software had to be
> listening to a particular port in order to even get the "Host"
> header, or anything for that matter.
Actually, this may not be true even for present software -- consider a
server running on a Unix system under inetd.
A small worry I have is that a "Host" header may be relevant only for
certain implementations of HTTP -- those using IP as an underlying
transport mechanism. ("Port number" information, of course, restricts
things even further.) I trust that the wording of the HTTP/1.1
specification will not make it impossible to implement the protocol
elsewhere.
Draft 03 of the HTTP/1.0 spec, section 1.2 paragraph 2, has a very
nice statement of the transport-independence of HTTP; I'd like to see
it stay that way.
For versions of HTTP beyond 1.1, it seems to me that a better
long-term solution might be to allow (or require) the "Request-URI"
for a request to an origin server to be an "absoluteURI". (This would
decouple HTTP from TCP/IP at the cost of tying it more tightly to
URIs, I guess.)
Cheers,
Andrew.
--
Andrew McRae <andrew_mcrae@harvard.edu>
Received on Saturday, 7 October 1995 13:11:18 UTC