Re: The HTTP 'link' and 'Content-Base' Headers

In order for items in a protocol to advance to Draft Standard, there
must be at least 2 tested interoperable implementations of both client
and server.  By IETF rules, such items must be removed for draft standard,
and there was no claim last summer that people would get them implemented
and tested.  There were no complaints at the time about this action,
nor were there promises of quick implementation.

Neither of these made the cut, at the time (late last summer).  See detailed
comments below.

The way to make progress on these in the IETF, is to write a separate (short)
internet draft defining them, show that you have the implementations (client,
server, and proxy (if there is any proxy specific behavior),
and they can/could/should progress (given that the headers were in the
proposed standard).  Note that the working group does NOT have to be
"open" at the time; the mailing list (will) continue to exist, and
the only difference in procedural rules is that there would have to
be a one month last call rather than a 2 week last call.  So I recommend
submitting such a draft, and it can be acted on presuming general agreement.

Default-Style is something I've NEVER heard of.  You can write an
ID for it, but do so separately, so that it can be acted on independently
(and not affect the other headers procedurally).  It could progress to
proposed standard in a similar fashion.
			Hope this helps,
				- Jim Gettys




> From: Ian Hickson <py8ieh@bath.ac.uk>
> Resent-From: Andy Norman <ange@hplb.hpl.hp.com>
> Date: Wed, 24 Mar 1999 18:28:42 GMT
> To: http-wg@hplb.hpl.hp.com
> Cc: connolly@w3.org, "L. David Baron" <dbaron@fas.harvard.edu>,
>         James Clark <jjc@jclark.com>
> Subject: The HTTP 'link' and 'Content-Base' Headers
> -----
> Hi.
> 
> It has been brought to my attention that the HTTP 'Link' header has been
> removed from the latest version of HTTP 1.1 (IETF's Internet Draft dated
> November 1998).
> 
> I understand that the reason for this change was that the HTTP WG
> considered the semantics to have been poorly defined. For example, one
> post on HTTP-wg says:
> 
> : There are known problems with some of the proposals: e.g. Link sytax
> : does not quite seem to be what the HTML group needs for enabling
> : server setting of sytle sheets; its grammar is suspect. [1]
> 
> Could you please explain what is "not quite what the HTML group needs"
> and why RFC 2068's definition of the grammar is 'suspect'?
> 

I don't remember the exact details; maybe Dan Connolly does...

In any case, as I remember we were left with two definitions: one in the 
HTTP spec, and one in the HTML spec, and without knowledge of which was 
exactly correct.

Furthermore, one needs 2 interoperable implementations to progress to
draft standard; at the time the interoperability reports were being
done last summer, there weren't known to be two clients, servers and
proxies.


> NGLayout, the engine behind version 5 of Netscape's browser, currently
> comprehensively supports the HTTP Link header, as defined in RFC 2068.
> No particular difficulties in interpreting the semantics of the 'Link'
> header were hit when working on getting this feature implemented.

Great.

> 
> This header is _incredibly_ useful, for, eg, setting the webmaster
> email address of every document in a site or for changing the
> stylesheet selected in every page of a 2000 document site, without
> having to edit each file individually. (The "Default-Style" header is
> very useful in this respect too.)

Never heard of a "Default-Style" header. What is it?

> 
> I have also noticed that the 'Content-Base' header was also deleted.
> NGLayout also supports this HTTP header.
> 
> Your time is greatly appreciated.
> 
> -- References --
> [1] quoted in:
> http://hplbwww.hpl.hp.com/people/ange/archives/archives-97/http-wg-archive/160
> 3.html
> 
> Test cases exist for both of those headers at
> http://www.bath.ac.uk/%7Epy8ieh/internet/importtest/
> 
> NGLayout latest build is available from
> ftp://ftp.mozilla.org/pub/mozilla/nightly/
> 
> Note that NGLayout currently only supports HTTP headers specified in
> META elements. This is due to a bug that is preventing any HTTP header
> from reaching the document's DOM, and will be fixed prior to release.
> 
> --
> Ian Hickson
> U+2642 U+2651
> U+262E U+2603 U+263A

Received on Monday, 29 March 1999 06:30:11 UTC