Re: Why do we have a component model?

On Mar 8, 2005, at 7:29 AM, <paul.downey@bt.com> wrote:

>> I am not yet convinced that we need to make drastic changes like this
>> instead of fixing bugs at this point. Component model provides an
>> abtraction to talk about composition within the language (i.e. WSDLs),
>> composition with another (i.e. Schema), composition with  
>> constraints...
>
> seems to me that we spend most of our time trying to fix bugs in the  
> component
> model,

Paul, you know I'm sympathetic to anti-component model claims, but they  
*must* be backed up. Roberto's "touching 7 places" claim was  
compelling....except that when he listed them out fewer than half were  
due to the component model (not as compelling).

So, I'd like some grounding of your "seems". A first approximation  
would be the number of issues or LC issues that dealt with this.

> in particular the area of composition.

Composition is tricky.

> The engineer in me wants to find a
> simpler solution rather than continue to add more sticky tape and  
> chewing gum.

That's an unfair, and derogatory, characterization. Composition *is*  
tricky and there is no evidence that I see that the component model  
makes that *harder* than it is and some evidence that it makes it  
easier.

And people seem to want composition (whether they require *our*  
composition....). E.g., the editor of the SPARQL protocol document, my  
good friend Kendall Clark wrote:
	"This idea of multiple interfaces and interfaces that inherit others
	that we get from WSDL is actually *incredibly* useful from a spec
	writer's point of view since it means we don't have to get into
	conformance levels and the like. You implement the interfaces you care
	about and you describe the ones you implement."

<http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JanMar/ 
0271.html>

> i'm drawn to the idea of spec which is focused on the document rather  
> than the
> processing model, even if that meant losing import (but keeping a  
> lexical include).

I do wonder if xinclude could be useful :)

There is the problem of composing with Schema too. You could refer to  
all schema types by their component designators, but you probably  
wouldn't be able to use qnames then.

If we can really find consensus for trimming features, great, but the  
record has shown that "WSDL 2.0 --- the Tom Jordahl Cut" is a flop for  
the test audiences, however much it is the darling of the critics.

Cheers,
Bijan.

Received on Tuesday, 8 March 2005 14:15:20 UTC