W3C home > Mailing lists > Public > xml-dist-app@w3.org > January 2002

Re: Article: Fat protocols slow Web services

From: Mark Nottingham <mnot@mnot.net>
Date: Wed, 9 Jan 2002 11:00:50 -0800
To: Mike Dierken <mike@dataconcert.com>
Cc: "'Xml-Dist-App@W3. Org'" <xml-dist-app@w3.org>, "'David Orchard'" <david.orchard@bea.com>
Message-ID: <20020109110043.D1183@mnot.net>


HTTP is not the place to do reliability. HTTPR is 'layered' on top of
HTTP in the sense that it defines a message format that is
transported in the HTTP body, making it opaque to HTTP, but it still
requires on HTTP. You can't take a HTTPR payload and stick it in
SMTP, for example.

If reliability is done this high in the stack, it should either
encapsulate the application protocol (a la TLS) or it should be
integrated well into the application protocol (e.g., as a SOAP
header, or much less preferably, an HTTP extension).

On Wed, Jan 09, 2002 at 09:23:33AM -0800, Mike Dierken wrote:
> > -----Original Message-----
> > From: David Orchard [mailto:david.orchard@bea.com] 
> > 
> > Why is there a need for reliable http?  I believe that 
> > reliability should be layered on top of HTTP, and SOAP is a 
> > good place to do it.
> > 
> > Imagine the multi-hop case, and you can immediately see why 
> > the higher layer needs end-to-end reliability.
> I agree. Has anybody done a message sequence diagram showing the http
> messages 
> that would flow in:
> a) multi-hope situation
> b) client initiated situation (the 'request' is in the response to a client
> request)

Mark Nottingham
Received on Wednesday, 9 January 2002 14:30:18 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 23:11:45 UTC