Re: GRDDL and OWL/XML

On May 12, 2008, at 10:21 PM, Jeremy Carroll wrote:

> I've read the thread so far, and only feel minded to respond to one  
> of Bijan's earlier points that hasn't got much attention.
>
> Overall, I'm sympathetic to Bijan's perspective: any W3C  
> implementation of the transform should in my mind clearly be  
> labelled as inferior and derivative to the normative spec; but I  
> don't think that means that there shouldn't be one. The practical  
> utility of its existence is the key argument for me.

I'm happy to discuss this consideration. I'm perfectly happy to  
concede up front that sufficient practical utility could mitigate  
downsides (even putting aside whether the downsides are as likely as  
I think they are).

(One problem with the OWL WG discussion of this is that it  
immediately shifted to a procedural basis, to wit, the charter  
requires GRDDL (actually, it doesn't) and GRDDL requires on line XSLT  
(also not required) thus we are forced down this path without having  
to consider the advantages/disadvantages. So, that tends to make me  
argue more strongly esp. to mitigate second order effects, to wit,  
having to have similar battles in the future. Frankly, I was pretty  
scarred by my WSDL/RDF experience. I think reaching out to the XML/ 
Web/Wider world requires a much more delicate touch that has been  
evidenced.  I do think in this circumstance the utility of a WG  
sanctioned reference implementation is very low, esp. when compared  
with the other work we're trying to get done. Even if I (and people  
like Boris) don't end up *doing* the implementation, our  
organizational duties would almost certainly compel us to do a  
detailed technical review of something which, at best, is a fairly  
marginal value *as a WG deliverable. This can be debated and I'm   
happy to debate it.)

[snip discussion of alternative GRDDL vision...I like it too!]
> I think the role of W3C blessed implementations is not as  
> 'reference' implementations, but seed corn; and I don't think that  
> your fears that the seed corn will be a weed stifling the growth of  
> better implementations is well-founded.

Could be. Though I'm not sure exactly what the difference between  
seed corn and a reference implementation is if it is blessed. Why,  
instead of blessing, don't we just maintain a list of  
implementations? Instead of going for a deliverable, just solicit  
implementations during CR? You could even put the list on the  
namespace page and let implementations (optionally) use whichever  
they prefer or let them ask the user.

If we didn't have easily accessible converters widely distributed  
(and available as restful web services) the situation might require  
seed corn. But to expend resources to by seed corn when we've got a  
nice harvest to thresh seems unnecessary. Then my point about  
competition can be pretty weak and still dominate.

Cheers,
Bijan.

Received on Monday, 12 May 2008 23:43:42 UTC