Re: Profiles intro

On 10 Apr 2008, at 14:50, Alan Wu wrote:

> Bijan,
>
>>
>> Ok, I'll rephrase your point: Regardless of whether it's true or  
>> not, RDFishness is not helpful in distinguishing fragments and may  
>> cause extreme negative reactions. Thus, we should find other  
>> points for guidance.
>>
>> Cheers,
>> Bijan.
> Could you please elaborate on the extreme negative reactions?

It's not saying that OWL-R *is* RDFish, but rather than the others  
are not.

Isn't this exactly Ivan's reaction?

> Triple-friendly is indeed one of the features of OWL R.

Actually, I don't agree, per se. At least not very interestingly so.  
YYMV. My point is that highlighting this can be contentious. It  
already *was* contentious. See this thread.

> This allows
> a plain triple store vendor to adopt/add some OWL easily. And I  
> don't see any harm in that.

You've got it wrong way around. See above.

> Of course, I don't intend RDFishness to be *the* distinguishing  
> point or *the only" distinguishing point. OWL R, as designed, does  
> facilitate/encourage
> implementations to materialize inference graph and consequently  
> speed up query execution.

That's an implementation point. Personally, I think materialization  
is a fairly bad implementation strategy (and yes, I have tried it).  
See various recent papers on RDFS reasoning.

OWL-R doesn't force materialization. It works just fine with a magic  
sets/semi-native evaluation strategy. But now we're off point :)

Cheers,
Bijan.