Re: RDF as a syntax for OWL (was Re: same-syntax extensions to RDF)

I see there is an olive branch at the end. I preserve my remarks along 
the way because I do not believe them defensive or unjust. But here I 
extend a Group Hug.

On Jan 6, 2005, at 7:24 PM, Jeen Broekstra wrote:

> I'll try and keep this short.
>
> My earlier messages were simply aimed at finding out to what extent
> you had tried to use the methods I suggested,  and what the exact
> problem was that you were trying to solve (which wasn't completely
> clear to me from reading the earlier posts).

There isn't "a" problem. There's building ontologies in, tools for, 
applications using, and extensions of OWL (and RDF), plus teaching and 
explaining it to people. I've not found it useful for any of these, and 
typically not neutral for them either.

> Your somewhat

Only somewhat?

> vehement
> dismissal of the triple representation of OWL seemed unnecessarily
> generalized to me,

No, it is quite unqualified. Or rather, it is quite qualified: The 
requirement of same-syntax extension has proven to have no advantages 
and many disadvantages in a wide variety of tasks and purposes. From 
specification to use. Species detection and nnf are just two, 
relatively simple, but real examples where the task is, given a decent 
representation, pretty trivial and given triples, rather ridiculous. 
They are just obvious examples, though. In return, I've heard *nothing* 
about the real benefits, aka "niceness", of the triples encoding.

Note I didn't sat it was *impossible* to do the tasks I mentioned, just 
that it was obviously much harder and cumbersome. Of course it's 
possible, we've done it. This spreads throughout most areas of OWL -- 
the triples represention is typically a lot of pain.

>  perhaps I have read that wrong however. If your
> only point is that for some tasks (like nnf conversion and species
> validation), triples are an unwieldy representation format, and that
> other formats are much better suited for these tasks, then fine,

*Most* tasks. And dude, thus far I've come up with killer examples, and 
you've come up with nothing. Show me an implementation task that's only 
a *little* worse using the triple representation.

>  I
> will happily accept this, especially since you indicated that the
> approach I suggested was actually more or less what you tried and it
> didn't work.



> A few last clarifications on my part:
>
> Bijan Parsia wrote:
>
> [snip]
>
>>> Then perhaps my problem with all this is that you seem
>> I humbly request that you stop with the seem claims. Either amass
>> enough confidence from your reading of my text to make a defensible
>> claim or ask me a question.
>
> Stop being so defensive.

Stop being so offensive.

> I'm not accusing you of anything here (nor
> anywhere else, for that matter). When I say "you seem" I merely mean
> "you seem to me", i.e. "I understood from your posts". I'm not putting
> words in your mouth, I'm merely trying to understand what the heck you
> are saying, and inviting you to correct me if I'm wrong.

I understood the rhetorically sleazy move you made with your weasel 
words. *Your* defensive reaction indicates that, whether you realize 
it, you weren't acting in good faith. Or, if you prefer, you certainly 
haven't communicated that good faith to me. If you have such trouble 
discerning the situtation, then perhaps you should try a different 
interpretation strategy.

What's so defensive about my asking you to put a bit more effort into 
your interpretation *or* start asking actual questions? *Hmmmm*?

> [snip]
>
>> Look, if you already  have a species validator, you don't need 
>> (necessarily) to write one. If you use the OWL API structures, you
>> don't have to think about triples. But the claim in question is
>> whether triples are *nice* to work with. My counter claim is that
>> my experience is that they are neither standalone nice, nor nice by
>> comparison, for a wide range of common tasks.
>
> As I said above, I think I can almost completely agree with this.
> Except that I have the feeling that you overstate the case when you
> say "wide range of common tasks".

Oh yes, you "have a feeling", or you "believe", or you "have an 
opinion". Based on what? You might find my claims inherently 
implausible but:
	1) there is some consensus of the experienced; of the implementers and 
specifiers and users who've posted, only Jos has claimed experiential 
counterevidence; that is, of course, an ongoing challenge to my view; 
but as the discussion has noted, he's lacking certain experience 
(though he is working on rectifying that -- we'll see what his judgment 
is afterwards)
	(Technically, sandro has also counterclaimed, but he's been silent for 
a while and only made the *bare* claim. Until he rejoins with some 
evidence, I think it fair to discount his *raw* assertion.
	However, he did do a translater to Otter format. We would need to 
examine that, perhaps in contrast with, say, the Manchester TPTP 
translater used in their Vampire work)
	2) I've given plausible, real world, easy to understand examples; 
Geoff has taken the nnf challenge, and I'll be following up to him 
shortly. *There* is a case of someone acting in good faith. He's 
perhaps not ready to concede the *level* of pain I claim, but I hope, 
in discussion, to exhibit the problems and pains the triples approach 
produces.
	3) oh, why bother.

> But I guess it's a matter of
> perspective,

Yes. Experience and understanding vs. speculation and faith.

> and I certainly concede that for at least the tasks you
> mentioned as examples, this is probably true.

Oh, how gracious of you.

> I still fail to see, however, how, in the wider scope of things, this 
> is a problem, since we know that other formats exist, to which we can 
> easily translate OWL ontologies, and that these other formats make 
> these tasks so much easier. You mentioned "playing the semantic web 
> game", but I don't believe that that necessarily means you have to 
> manipulate triples.

Well, look. The Semantic Web is, in part, a standards and 
interoperability game. The discussion here is directly targeted at 
figuring out what kinds of objectives and requirements a future working 
group (perhaps starting from SWRL) would have. Sandro is trying to 
determine whether a same-syntax semantic extension approach is possible 
and desirable for SWRL/FOL semantic web languages as standardized by 
the W3C. My experience is that the enforcment of a same-syntax semantic 
extension on OWL, which there's some strong though exhibits the 
*limits* of the *possbility* of that approach, was a completely 
unnecessary burden with no benefits. Architecturally, it's misguided 
(as I pointed out, XML is *supposed* to be the syntax layer!) and based 
on severe misunderstandings. It's been a drag on tool development. It's 
a drag on education and adoption. It's a drag. It *makes work* where 
there is no need for such making.

Now, of course, I can work around it and I can move to other things. 
But if I want to work with semantic web data I cannot entirely ignore 
it. At least I must write bridge tools. I must explain things to 
people. Etc. etc.

Plus, it's a *lot of work* to specify the triple encoding, itself. 
Correctness is hard to get. *Sense* is hard to get! I understand you do 
not believe that. You are free to keep on disbelieving it. But I have 
what I consider to be very strong grounds for my views, both 
theoretical and experiential.

I'm not resting on authority there. I'm perfectly willing to back up 
each claim as we go. But I deny you the right to provisionally scope my 
claims until you have evidence in your favor. The burden of proof is on 
your side, not mine. (I've given several burden shifting arguments 
along the way, so it's not a brute arrogation of dialectical position 
on my part.)

> [snip a lot of confusion and misunderstanding from both of us]

I deny the latter, fwiw.

> Finally: it was at no point my intention to ridicule you or your
> approach, merely to find out what the scope of your problem with RDF
> triples for OWL was. You seem to have taken offense at this however,
> for which I'm sorry.

Accepted.

It might be helpful if we would establish what counts as evidence in 
support or against a claim, and to deal with specific claims in turn.

Cheers,
Bijan Parsia.

Received on Thursday, 6 January 2005 12:48:38 UTC