Re: content negotiation anti-principle

On Thursday 09 January 2003 11:35 am, Jeremy Dunck wrote:
> Gavin wrote off-list, and I'm replying (with his permission) on...

I'm just going to reply to this message, rather than all of the recent 
messages.

> I replied previously [1] to a similar statement.  It's my feeling that
> bringing up negotiation into the markup would actually raise the cost to an
> author in the general case.

This is true... you shuffle some of the pain upward, but only in the cases 
where the author cares about it. 

> That sounded stronger than I intended.  What I mean is, I think there is
> a chance for server-side negotiation to be cheaper (causing more usage of
> it), without significantly impacting the existing web.

I think this is true, but it only solves the problem for people managing 
servers. Again, there are two problems: the difficulty of deploying content 
negotiation (which leads to it not being deployed much), and the fragility it 
causes for applications.

> Changing the markup as a primary means of addressing the problems with
> server-side negotiation seems to be forcing the world to change its
> authoring habits, and still places the burden of decision on the server.
> That doesn't seem like a success, to me.
>
>   Can you help me understand better?

I'm not saying that moving content negotiation knowledge upward solves the 
problems of server-side deployment. That needs to be solved with better 
tools.

Giving authors control helps because there are a number of cases where the 
specific representation returned can have a significant impact on the ability 
for the application to process the resource... sometimes fatally so. This is 
more of an issue with XML media types than images, because the set of 
variants is huge, and dynamic.

Again though, you have to ask whether it's really worth it. Elliot made the 
comment that most parsers asking for an entity that uses an http URI SOI will 
ask for text/xml or application/xml. This is an example of where there are 
specific rules implied from the context of the URI, and these kinds of rules 
should, perhaps, be codified. If they were, I think much of the need for, and 
many of the problems with content negotiation might well go away.

BTW. Somebody pointed out to me, rightly, that this is really travelling the 
NOTATION path all over again...

Received on Thursday, 9 January 2003 22:46:27 UTC