W3C home > Mailing lists > Public > ietf-discuss@w3.org > January 1999

Re: HTTP Extension draft

From: Henrik Frystyk Nielsen <frystyk@w3.org>
Date: Mon, 04 Jan 1999 09:19:59 -0500
Message-Id: <3.0.5.32.19990104091959.03125100@localhost>
To: Rob Lanphier <robla@real.com>, discuss@apps.ietf.org
At 00:06 1/4/99 -0800, Rob Lanphier wrote:

Hi Rob,

>However, I can speak from the perspective of a vendor of an HTTP system who
>may be confronted with implementing this one day.  I think it solves the
>problems it tries to solve well, but that it bites off more than it should
>chew.
>
>Here's the three problems that I see this solving:
>1.  Ability to add mandatory extensions
>2.  Ability to send extension metadata
>3.  Ability to add vendor-specific extensions free from namespace collisions

Thanks for your review - first I have a meta question: In #2 what kind of
metadata do you mean? In case you mean the old feature discovery exchange
mechanism defined in PEP then I fully agree with you. If you mean the
extension instance parameters then I would not consider that metadata but
call parameters of a particular instance.

With respect to #3, is it that you consider a central registry sufficient
or is it the name space prefix model you are referring to? Note, however,
that as a policy neutral specification, there is nothing that prevents an
extension from being open or to be the design of multiple organizations
and/or commercial companies. This is fully up to the designers of that
particular extension.

The main purpose is to define a mechanism that allows for graceful
deployment of new extensions locally as well as globally without requiring
that two peers have complete knowledge about each others capabilities.

>So, in the best possible of worlds, I would prefer to see proposals that
>decouple the three, and solve them semi-independently (though #2 and #3
>could certainly build on #1).  I'd be very supportive of a proposal that
>solves #1 as simply as possible, and tests the limits of that model.  #2
>and #3 could be saved for HTTP-NG, or could be added if the #1 solution was
>proven insufficient.

The HTTP extension draft in fact does separate instance data from metadata
- it simply doesn't define a mechanism for handling the latter
independently of instance data.

Thanks!

Henrik
--
Henrik Frystyk Nielsen,
World Wide Web Consortium
http://www.w3.org/People/Frystyk
Received on Monday, 4 January 1999 09:20:45 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 23 March 2006 20:11:25 GMT