Re: Extensibility in Syndication formats

[Bill de hÓra <bill@xxxxxxxxxx>]

Dare Obasanjo wrote:

> How is this practically useful? The fact that an
> extension will show up as single elements with simple
> content or as elements with attributes and complex
> contents doesn't bring my application any closer to
> understanding them when encountered in the wild. 

Speaking as an avowed and somewhat loudmouthed RDF cynic.

It's useful because one can add new data items to markup with 
breaking existing code downstream. Whether you want to add code to 
meaningfully process those data items is another matter. You seem to 
be deliberatly mixing those two things up, or perhaps you don't 
understand the difference. In my experience XML+namespaces doesn't 
provide anything like that. Quite the opposite it's sufficiently 
fragile to almsot assure breakage down stream and down time, so much 
so entire communities are encouraged to figure out a sensible means 
to layer extensibility onto schema driven XML documents, something 
that we'll be doing here soon enough no doubt. Much of this 
fragility I see as I suspect is a direct result of people believing 
that XML+namespaces have some magical properties of extensibilty and 
getting bitten later on.



> So far you have listed a bunch of drawbacks of the RDF
> based approach used by RSS 1.0 and none of the
> benefits that haven't been seen using just XML and
> namespaces in RSS 2.0. 

That's makes so little sense it's not even wrong. When it comes to 
extensibility or flexibility in markup, making these sorts of claims 
around "just XML and namespaces" is bunk. For example, I've read 
yours and others useful articles and papers on extensibility. Yet 
that's not just XML and namespaces, that's an entire added semantics 
on top "of just XMl and namespaces" *and* schemata. So really you're 
comparing RDF/RSS1.0 with "just XML and Namespaces" and schema and a 
lot of other articulated processes and constraints. We here, already 
have XML+namespaces for Atom, remind me again why we need a model 
extensilbilty?

In OO terms, the difference between RSS1.0 and all the others is 
very like the difference between programming with 
reflection/hashmaps and programming with domain specific objects. 
Both have their uses, bth have their drawbacks. The primary tension 
in my experience there is a balancing act to conduct between the 
flexbility provided by something like RDF versus the clarity of 
domain specific markup. Raw RDF is not neccessarily comprehensible 
with respect to a 'domain model', raw XML+Namespaces is not 
neccessarily flexible with respect to 'change'. I suspect it helps 
to have tried both approaches rather than simply looking at them and 
opining to appreciate this.

cheers
Bill