Re: draft public comment for OWL last call from RIF

I see.  All this because I said "ridiculous".  A poor choice of words, I 
apologize.  Here:

Stepping back from RIF and OWL, it would be a shame if they each supported a 
different set.

-Chris

Bijan Parsia wrote:
> On 23 Jan 2009, at 14:20, Chris Welty wrote:
> 
>> Bijan,
>>
>> While RIF and OWL are clearly different, and have different goals, 
>> they are also very similar, and there is a significant portion of the 
>> world that views them that way.
> 
> That's fine. All I'm arguing is that it's not *ridiculous* (your term!) 
> for there to be divergence in the *set* of datatypes supported.
> <http://www.w3.org/mid/4978939C.3000900@gmail.com>
>     """Stepping back from RIF and OWL, it seems ridiculous to me that 
> each would maintain a different set."""
> 
> "Ridiculous" means there's no case to be made. I believe there is a case 
> to be made (depending on the circumstances). There's also a case to be 
> made against.
> 
> One could make a better case that difference in the *semantics* of same 
> named datatypes should be the same. But even there, OWL does *different 
> things* with the datatypes (i.e., it treats them as constraints).
> 
>> (Recently, giving a RIF talk, I was asked "So which do I use, OWL or 
>> RIF?"  I did not perceive from the rest of the audience any indication 
>> that this seemed a ridiculous question.)
> 
> I don't see that it's a ridiculous question even if the datatypes 
> diverge. One possible answer would be "OWL supports owl:rational, so if 
> you need that, either work with OWL or an extension of RIF."
> 
> I don't see that that's a ridiculous answer.
> 
>> So while you see no strong argument to have the same set of datatypes, 
>> I see no strong argument *not* to make them the same,
> 
> The strong argument is that the OWL community wants, afaict, more 
> datatypes than, from what you seem to indicate, RIF implementors want to 
> support. Since I'll bet that *all* systems that will use RIF for 
> interchange that are outside RIF, I don't see why OWL has to be held to 
> the lowest common denominator.
> 
>> quite the opposite in fact - it seems plainly obvious to me they 
>> should be the same.
> 
> "Plainly obvious" is not an argument of *any* kind.
> 
>> You have given no strong argument here, other than your own personal 
>> motivations.
> 
> How are my "personal" motivations as a prominent vendor and users of OWL 
> systems irrelevant? It's certainly more relevant than bare assertions 
> otherwise.
> 
> If you like, I hereby say that the University of Manchester has an 
> interest in a expansive set of datatypes to be supported in OWL 
> including owl:rational. See the *extensive* discussion about it in the 
> OWL archives.
> 
>> And I do not think it futile
> 
> I think it's futile because I think, and plan, to add a number of 
> additional datatypes to OWL For example:
>     http://www.webont.org/owled/2008/papers/owled2008eu_submission_44.pdf
> which has a lot of support (including from the Product Modeling XG).
> 
> I suspect that some RIF based vendors' user bases do not need such 
> types, thus I'd be surprised if they would get into the "common 
> denominator" of RIF. But there's no reason, therefore, for them  not to 
> be in the common denominator of (a future) OWL.
> 
> Interchange between widely disparate systems often is well served by 
> restrictions. User modeling often is not. Hence, the futility.
> 
>> - if you believe its non-futile
> 
> I don't believe it is non-futile. Or, more to the point, I don't think 
> it's necessary. Without *some* argument for why they being the same 
> (esp. as extant SW languages such as RDF and OWL 1 haven't had the same 
> set) is an overriding consideration, I don't feel any extra burden of 
> proof at all. I have plenty of arguments for the datatypes Manchester 
> wants. I don't presume to tell RIF vendors what datatypes their users 
> want so accept their arguments that they want fewer datatypes. Thus, we 
> have a divergence. If all other things were equal, I'd happily 
> synchronize them, but the burden of proof is that all other things are 
> equal. You are trying to shift the dialectic such that divergence is not 
> just prima facie wrong, but requires heroic arguments to overcome. I 
> don't see that that's a shared presumption. I don't share it. OWL 1 and 
> RDF WGs didn't share it.
> 
>> and should be the job of one working group to formally designate a set 
>> of datatypes, then it is certainly possible for two working group to 
>> do it, albeit the amount of work increases by an order of magnitude 
>> which I don't relish.  None the less, I see this as our responsibility.
> 
> I'm happy with having divergent sets of datatypes *within OWL Profiles* 
> though I do prefer, all things being equal, for them to be the same. But 
> I'm happy to consider, without prejudice, implementation and user base 
> considerations.
> 
>> We do seem to agree that the set of datatypes should have the same 
>> semantics.
> 
> Up to a certain point, yes. E.g, I wouldn't want to force RIF 
> implementations to do CLP by default.
> 
>> This makes me think I should send two message from RIF, one on the 
>> different semantics and one on the common set of datatypes.
> 
> I would recommend that. I would also recommend that you don't presume 
> that everyone agrees with what you find obvious or is ridiculous.
> 
> When it comes down to it, as a user, I don't care about owl:rational in 
> RIF, but I really care about it in OWL. That's why I didn't make an LC 
> comment to RIF about the lack of owl:rational.
> 
> Cheers,
> Bijan.

-- 
Dr. Christopher A. Welty                    IBM Watson Research Center
+1.914.784.7055                             19 Skyline Dr.
cawelty@gmail.com                           Hawthorne, NY 10532
http://www.research.ibm.com/people/w/welty

Received on Friday, 23 January 2009 15:29:06 UTC