W3C home > Mailing lists > Public > ietf-discuss@w3.org > November 2001

Re: Requirements for reliable message delivery

From: Mark Baker <distobj@acm.org>
Date: Fri, 30 Nov 2001 01:10:03 -0500 (EST)
Message-Id: <200111300610.BAA03682@markbaker.ca>
To: moore@cs.utk.edu (Keith Moore)
Cc: discuss@apps.ietf.org
> > HTTP's application model is not at all restrictive.  IMHO, it's the
> > most powerful and generic one ever created.  I wrote about this
> > recently;
> > 
> > http://www.markbaker.ca/2001/09/draft-baker-http-resource-state-model
> 
> I like your description, but I fear it completely misses the point.  

Yes, it doesn't come right out and say how generic the semantics are.
I was hoping that would be inferred, but so far nobody's picked up on
it. 8-(  I could add that as an appendix, I suppose.

But let me write it here quickly ... the application semantics are
equivalent to containment based software component models, which are
extremely generic and expressive.

> Second, the way that HTTP POST is typically used (and the way in which 
> protocol authors want to use it) bears little resemblance to that
> described in your draft.

HTML FORMs are the most common use of POST, and that use is consistent
with POST semantics, as I attempted to describe in my draft.

But I agree that there's lots of misuse, such as RPC-over-POST,
IPP, etc..

> Third, you have failed to support the argument that even the state model 
> you describe in your draft is either 'powerful' or 'generic'.  

See above.

> Most people who want to use HTTP aren't attracted to its state
> model. 

No argument from me there!  But I think if more of those people knew
what could be done using only HTTP as it was designed to be used,
there'd be less reason to use it improperly.  That was the primary
motivation for my draft.

> They're attracted to HTTP because it runs over port 80, works
> through firewalls (usually), seems to have security (which may or 
> may not be appropriate for the user's particular purpose), can be 
> plugged into an existing server, and/or (mostly) because they are 
> (somewhat) familiar and comfortable with it.  They (often incorrectly) 
> assume that by simply using HTTP they can avoid worrying about protocol 
> design issues.

Yup.

MB
-- 
Mark Baker, Chief Science Officer, Planetfred, Inc.
Ottawa, Ontario, CANADA.      mbaker@planetfred.com
http://www.markbaker.ca   http://www.planetfred.com
Received on Friday, 30 November 2001 01:12:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 23 March 2006 20:11:29 GMT