why we need GRDDL (Re: Grddl et al)

> as an alternative approach, why don't we make a
> 
> - call for implementations (in the candidate recommendation phase) and
> - gather pointers (including a pointer to a GRDDL document)
> - and then summarize these in a document (the CR implementation report).
> 
> This report might eventually be out of date, but then so be it.
> 
> What would be wrong with this?

If GRDDL works on day one, then systems consuming OWL (in the RDF/XML
serialization) with a GRDDL processor in the loop can keep consuming all
W3C Recommended OWL.  If GRDDL doesn't work on day one (and keep
working), those systems will be cut off from the new world of OWL using
the XML Serialization (unless the implement a new parser).  Those folks
(the people who have adopted OWL so far) will be justifiably unhappy
about this.  Othor people may also be pressured not to adopt the new XML
Serialization because their ontologies will be unusable by some
significant fraction of the OWL world.  (This is in addition to the OWL
1 vs OWL 2 fracturing, that's a necessary cost we're planning for.)

Now, it may be that in practice no one consumes OWL via RDF/XML with
GRDDL in the loop.  It may be that, in practice, folks consuming RDF/XML
are faced with the option of either implementing GRDDL or implenting the
new XML Serialization.  In that case, they might well skip GRDDL.  That
would be okay, I suppose, but GRDDL will (in theory) have many other
uses as well, so they probably should implement GRDDL soon anyway, for
loading their data from their XML sources.

Procedurally, as Alan mentioned on yesterday's telecon, the commitment
to deliver GRDDL has already been made [1].  People can argue what
exactly the commitment meands (eg Bijan's point that it doesn't
necessarily mean XSLT or on-line), but I think people reading the
charter could reasonably understand us to be promising a working on-line
XSLT-based transform -- that's certainly how I read it.  I think any
attempt to back out on that deliverable would need to be renegotiated
with the broader community (technically via a re-chartering, but a broad
discussion across the Semantic Web community would probably be enough).
Basically, if we're going to break a promise, we have to make sure no
one will mind, right?

So, is there some big problem with writting the XSLT?  I suspect soon it
will become more cost-effective to implement it than to keep discussing
it.

       -- Sandro

[1] http://www.w3.org/2007/06/OWLCharter.html#deliverables

Received on Thursday, 8 May 2008 15:54:38 UTC