Re: Can we ensure that Content Negotiation works?

[Danny Ayers <danny.ayers@xxxxxxxxx>]

On Wed, 22 Sep 2004 20:19:25 -0400, Joe Gregorio <joe.gregorio@xxxxxxxxx> wrote:
> 
> My experience says to avoid relying on content negotiation
> in web services:
> 
> http://wellformedweb.org/news/WebServicesAndContentNegotiation
> 
> Not to say that you shouldn't send a proper Accept header
> with every request, just don't consciously place multiple
> types of resources at the same URI.

As a pragmatic general rule-of-thumb this seems good advice. However,
I note that virtually all the problems you identify (and Norm
identifies) relate to bugs in existing software. Given that the Atom
spec will control both the horizontal and vertical (or rather server
and client) of the situation, I don't think current bugginess is
justification enough to avoid conneg if it offers a good solution for
correctly-functional systems. I don't see any significant anti- case
in the Web Arch. doc either.

But the example you give where you are passing URIs to a service and
lose control of the negotiation is a killer. If this is a likely
scenario in the use of Atom, then conneg will simply be impossible to
use in the general case.

Do you think a setup like the W3C's XSLT service is a likely scenario?

Cheers,
Danny.

-- 

http://dannyayers.com