W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1996

Re: suggested wording concerning Host

From: Koen Holtman <koen@win.tue.nl>
Date: Tue, 2 Apr 1996 23:08:22 +0200 (MET DST)
Message-Id: <199604022108.XAA14121@wsooti04.win.tue.nl>
To: Dave Kristol <dmk@allegra.att.com>
Cc: fielding@avron.ICS.UCI.EDU, jg@w3.org, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Dave Kristol:
>  > > DMK:
>  > > I've noticed a cumbersome locution spreading through the HTTP/1.1
>  > > draft that I would like to cut off.  Here's an example:
>  > > 
>  > >     ... and the Host request header (present if the request-URI is not
>  > >     an absoluteURI) ...
>  > 
>  > Roy:
>  > Argh, where did that come from?  The Host header is ALWAYS present in
>  > HTTP/1.1.  It is never removed, not even when the full-URI is present.
> > It will not be removed until HTTP/2.0, which is a different specification.
>
>The words were in Koen Holtmann's mailing on content negotiation.  I
>agree with your point that Host is always required, so that's an error
>in Koen's wording.

Yes, the addition of `(present if...)' seems to be an error in my
wording.  I must be slightly out of sync on the host issue: I thought
that a user agent could omit the Host header when talking to proxies.

>But it's the intent of the wording I'm trying to get at.  There will be
>other places where someone is trying to say, "the host that appears in
>absoluteURI, or, if there is no absoluteURI, the host that appears in
>the Host request header."  I want to shorten that locution to
>"request-host".

I just went through the old 1.1 draft, and there are plenty of other
places where someone is trying to say this.  The current language used
is `the resource identified by the request-URI'.

The root of the problem seems to be that the language `the resource
identified by the request-URI' is inaccurate.  It has always been
inaccurate (the resources identified by http://a.com/index and
http://b.com/index are different, but have the same request-URI in
request messages to origin servers), but this inaccuracy has become
too great to endure in a web where we have vanity hostnames mapping to
the same server.

I do not think that changing

 `the resource identified by the request-URI'

to

 `the resource identified by the request-URI and request-host'

is radical enough.  It has some problems, because sometimes, the
request-host must be ignored.  After a close look at the current 1.1
draft, I have concluded that we can solve the problem by changing all
existing occurrences of

 `the resource identified by the request-URI'

and my new

 `the resource identified by the request-URI and the Host
  request header'

to

 `the resource identified by the request'.

It would also be good to add explicit language about how a request
message identifies a resource to Section 5.1.2.  I have found plenty
of language there on how to make a request message if you have the
full resource URI, but not much about deriving the full resource URI
from an incoming request message.

I would be willing to write language for section 5.1.2, but I think
the Host issue owner is better qualified to do so.

>Dave Kristol

Koen.
Received on Tuesday, 2 April 1996 13:14:52 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:49 EDT