W3C home > Mailing lists > Public > public-owl-wg@w3.org > May 2009

Re: CR Exit Criteria

From: Bijan Parsia <bparsia@cs.man.ac.uk>
Date: Thu, 28 May 2009 12:03:43 +0100
Message-Id: <DC670048-EF0D-483C-BA5D-0DCB46844E34@cs.man.ac.uk>
Cc: "OWL 1.1" <public-owl-wg@w3.org>
To: "Michael Schneider" <schneid@fzi.de>
On 28 May 2009, at 11:04, Michael Schneider wrote:
[snip]
> So, to restate what I said before: I consider it to be *not* likely  
> that
> the (now approved) CR criteria for OWL 2 Full are possible to be met
> within the time constraints of the CR phase.

This just seems obviously false at least on straightforward readings.

Exit criteria are *guidelines* and they also are, as with anything,  
subject to *interpretation*.

I find it hard to imagine that we'll get any evidence that OWL Full  
is *less* implementable than it was in OWL 1. OWL 1 Full was deemed  
sufficiently implementable to go to Rec. I don't think we've either  
formally put tighter constraints down or informally.

> And to confirm what you say above: I don't know about any literature,
> I consider it hard and non-obvious to implement OWL 2 Full, and I
> also consider the uncertainty to be high. This is all what I was about
> in my last mail. I certainly wouldn't have written the mail if I would
> think that OWL 2 Full was easy and fast to be implemented.

Sure.

> Concerning what I said about "obviously un-implementable", you seem to
> interpret more into this notion compared to what I intended. I just  
> wanted
> to express that I cannot point to a specific feature of the  
> language that
> would make me say: "Forget about implementing OWL 2 Full!", which  
> would,
> of course, be a show stopper for having this language under the W3C  
> flag.

Not obviously. Lots of implementors could take reasonable look at OWL  
2 Full 1 or 2 and say, "forget about implementing that". But some  
parts are pretty trivial: Property punning and infinite OWL Thing are  
supereasy. I can do a pretty easy incomplete implementation of the  
Hilog entailment by a prepreocessor and a somewhat better one using a  
looping preprocessor to a fixed point. I can pun vocabulary. I can  
handle rdf:list (incompletely). Etc.

Indeed, I'm rather surprised you didn't either notice or point any of  
these things out. They are fairly obvious and have been discussed in  
this working group.

> Actually, I really believe that one can do significantly better  
> than, say,
> what Jena provides (without hunting for unachievable theoretical
> completeness, though).

Why do you think completeness is unachievable. That's much stronger  
than I would hazard.

> But it will take a lot of effort, and most probably
> some new approach to pursue. Well, the book has to be written first,
> before one can implement from it. :)
>
> Btw: What I said about "building DL from scratch" was (as in my last
> week's mail) meant for the imagined case that there would be
> *no* existing theoretical foundation available (no books to implement
> from).

If that's what you meant, then, afaict, you didn't say what you meant.

> My (maybe hidden) argument was that I do not see why
> the desire to have an ontology language for the SemWeb would
> necessarily be bound to existing research in the field of
> implementing knowledge representation languages.

For research, knock yourself out. For standardization, things are  
generally different. If you had more experience in the W3C you'd hear  
a lot of "no research" expressed about working groups.

> The WebOnt WG
> would have certainly be able to produce a sort of OWL even without
> such theoretic work.

With the high liklihood that the language would be unimplementable  
(in reasonable time), unusable, or both.

What's the point of this? I have no idea what you are trying to  
achieve with these musings.

> But I wonder what the CR criteria would have
> been in such an "unlucky" situation.

Does it matter?

Cheers,
Bijan.
Received on Thursday, 28 May 2009 10:59:40 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:42:12 UTC