profiles and the rec track

I've been staying pretty quiet about this whole issue, but have been  
getting a fair amount of mail from concerned people (and you've seen  
some of it reflected in the LC comments - but many of what people send  
me are more "meta" comments, discussing the overall design, not  
specific features)
  My feeling at the moment is most of the concern is in the area of  
the profiles -- but let me be clear, it is not with the specific  
design of the individual profiles (the area I have argued in the past)  
but more generally about the whole issue

  Here are summaries of the issues that have been sent to me by other  
AC members, asking my opinion
   1 - there's a lot of confusion about QL vs. RL -- I think this is  
more than just an issue of "explaining it right" - there's just not  
enough experience in implementations to be able to say what queries  
each really handles better in practice and what the tradeoffs are
  2 - there's a lot of question about EL vs DL - again, the primary  
explanations are with respect to theoretical performance, but the AC  
members aren't interested in that - as one pointed out to me "there's  
a lot off polynomial stuff that doesn't work in practice, and some  
exponential and undecidables that are in daily use" - he's right.
  3 - there's a fairly large conceptual step in understanding the  
differences from OWL to OWL2 - several AC members have suggested to me  
that if we were to move OWL 2 (DL and Full) out the door with no  
profiles, and then bring the profiles along later, we would probably  
get more short-term OWL 2 use and then people would be better ready to  
understand the profiles and their use.

Frankly, this third one is the one I find most compelling - the WG has  
done a lot of good stuff - but the problem is with so much of it  
coming out at once, we will create confusion and risk backlash.

I have talked to a number of people about this - Here is my reluctant  
suggestion (and please note, I am reluctant - I know how muchwork has  
gone in, and remember that I have sent out a number of things  
supportive of RL in particular)

RECOMMENDATION FROM THE AC MEMBER OF RPI
  I think the WG should seriously consider taking the profiles off of  
the Rec track for now, getting the rest through, and then putting the  
profiles either into a separate CR or leaving them as WG notes.

I believe this would lead to more, not less, of OWL2 - I think it  
would let the profiles have more time to be tested in industry  
settings (esp. with respect to scalability on the Web).  Since they  
would be well-defined, I think companies (whether C&P or Oracle) would  
still be willing to support them .

Also please note that most of the "not well formed" LC comments are  
about the profile issues, not the overall OWL2 DL design, and thus we  
could postpone those for now.

  I know most of you will have a visceral reaction against this  
recommendation, but I hope you will sleep on it -- ask yourself this -  
which is more important - that OWL2 get out there and OWL gets more  
use, or that the profiles in their current form are published

   yours reluctantly
    JH




On Jan 27, 2009, at 3:57 AM, Bijan Parsia wrote:

>
> I think it's a bit incorrect to characterize FH6 as "Too many  
> profiles (supporting Lilly's comment)". Indeed, Liily's comment on  
> this point is a bit incoherent...the profiles result from research  
> efforts but OWL 2 is difficult to find useful subsets of?
>
> I'm really not sure what the complaint is, but as it doesn't seem to  
> be actionable, I think we shouldn't break it out separately. Or at  
> least, be a bit cautious.
>
> Cheers,
> Bijan.
>

"If we knew what we were doing, it wouldn't be called research, would  
it?." - Albert Einstein

Prof James Hendler				http://www.cs.rpi.edu/~hendler
Tetherless World Constellation Chair
Computer Science Dept
Rensselaer Polytechnic Institute, Troy NY 12180

Received on Tuesday, 27 January 2009 15:48:00 UTC