Re: profiles and the rec track

Bijan absolutely correctly points out that I am being inconsistent  
with my earlier position that all the deliverables should end up rec  
track.  I freely admit that.  In this case, I'm coming at it from the  
point of adoption of the language (and the AC politics)  - my fear is  
simply that the confusion over so many profiles will not be good for  
OWL - and rather than arguing for or against any particular ones, I  
think we should take the ones we have, make it clear in public that  
the designs are stable, but not move them to CR when we move OWL2 DL  
and OWL2 Full.  I believe that this would still allow all three of  
these important profiles to go forth in use, and interoperability  
issues would not be a big deal because the specs would exist - but it  
would allow more people to get used to them, to play with them, and to  
understand why the are important and when to use which.  However, it  
also means that confusion as to what "OWL" is and to why you might  
want to use it would be reduced.
  I would hold up as an example RDFS -- as painful as it was, RDFS  
didn't get to Rec its first time out -- however, as the design was  
mostly stable, it got heavily used among the early adopters, and  
supported by the first wave of RDF tools.  Some usages that hadn't  
originally been intended came along (such as having multiple domains  
and ranges) and these were addressed in the community (without needing  
the full mechanism of a WG to figure them out) and when RDF Core came  
along, RDFS moved out smoothly into a supportive user community.  I  
believe that OWL2 QL/RL/EL could go that same route - and if we make  
the decision to wait, we would avoid this decision possibly being  
asked of us during PR.
  Do I think we could get the profiles through the rec process? yes.   
But what I question is whether in doing so, we will have many people  
wondering, as Susie Stephens put it in her email:

> 1.  Given the usage of OWL 1.0 is quite limited
> in the industry compare to the usage of RDF, it may cost many extra  
> efforts
> and is very challenging to teach system developers to use new OWL2, in
> particular, identifying different subsets of OWL2 for developers with
> limited logic background


I think she is very perceptive, and I know my recent talks, where I  
try to overview the three subsets for web developers is very  
complicated (and sometimes hard to answer - i.e. for a particular  
application whether to go with RL vs QL, or EL vs DL, when the  
requirements are not fully known in advance) - in the industrial  
world, and especially in the government sector, confusion often leads   
to possible non-adoption (cf the scenario in the TQ LC) and I think we  
as a Wg should take that seriously

   -JH



On Jan 27, 2009, at 2:40 PM, Bijan Parsia wrote:

>
> On 27 Jan 2009, at 19:06, Jim Hendler wrote:
>
>> Bijan
>>  I will pass your email on -
>>  As far as charter, you yourself have said a number of times that  
>> it is up to the WG to decide what is rec track and what isn't
>
> I still believe that. I'm just wondering if you agree with the  
> advice with regard to profiles as you give with regard to UCR in:
>
> 	<http://www.w3.org/mid/C76F33CF-12C5-4998-9F2E-F70068195F9A@cs.rpi.edu 
> >
>
> That advice seems reasonable. It's not absolute, of course. But it  
> seems that we may be in for trouble *either way*.
>
>> - I think we've certainly met the charter requirements on the  
>> design of the profiles.  Yes, I have said at other times I'd like  
>> to see the deliverables rec track, and if it was a perfect world  
>> I'd still be arguing for the profiles, but as I said in the email  
>> that started this thread, this was a new realization for me, and  
>> took a lot of thought.
>
> I'm just concerned that we make our decisions in a consistent way.  
> One way of reading the charter is that the OWL 1.1 language (i.e.,  
> all the spec stuff) needs to be rec track. As part of that general  
> recommendation we must produce some informative stuff. If we unioned  
> all the documents into one big spec, we would go through and mark a  
> bunch of things non-normative. One way to do that is to put them in  
> a different document which is a note.
>
> Here we're talking about taking what was generally understood to be  
> a part of the language design and making *it* non-normative.
>
> I'm just pointing out that this is potentially procedurally non- 
> trivial, regardless of the merits. Of course, I'm often willing to  
> try non-trivial things if the merits warrant it :) But then the  
> arguments need to be pretty strong. Consider the last call comments  
> thus far on profiles:
>
> +0.1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/att-0051/index.html 
> >
> 	"""Profiles
> We are generally supportive, but would not oppose a scaling back of  
> this effort in light of comments from other consortium members."""
>
> +0.9 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0046.html 
> >
> 	"""and a very quick look at "Direct Semantics" and "Profiles". ...
> So, I think these docs are fine and the features introduced in OWL2
> interesting (though much anticipated)."""
>
> -1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0042.html 
> >
> 	(Frank)
> +1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0039.html 
> >
>       (no details, but I feel confident that he's supportive at  
> least of OWL RL)
>
> -1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0033.html 
> >
> 	Susie Stephens for Lilly
> +1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0025.html 
> >
> 	"""I¹d like to express my support for the proposed extensions and  
> profiles
> described in OWL 2 Web Ontology Language New Features and Rationale.
> This is from the perspective of implementing SNOMED-CT, where the EL  
> profile
> is particularly relevant."""
>
> +0.5 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0024.html 
> >
>        (No details, but discusses them.)
>
> +1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0020.html 
> >
> 	"""The documents on the different language profiles is extremely  
> welcome as
> are the different tractable fragments. For us the tractable fragments
> are extremely useful in providing guidance on which aspect of OWL 2 to
> use for certain problems. We expect to have many applications of OWL  
> at
> Ordnance Survey, and each of these fragments will be applicable for
> solving different problems. Also section 5 on the computation  
> properties
> was a very welcome and useful (and something I'd wanted to see for a
> while). This should make things clearer to Person and not Logician.  
> """
>
> +1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0019.html 
> >
> 	They are proposing adding stuff to OWL QL
> +1fuzzy <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0016.html 
> >
> 	"""I'd like to express my interest in and support for the proposed  
> extensions
> to be incorporated in OWL2, particularly from the perspective of  
> SNOMED-CT
> whose development I'm actively involved in."""
> 	(SNOMED people generally support OWL-EL
> +1 <http://lists.w3.org/Archives/Public/public-owl-comments/2009Jan/0014.html 
> >
> 	Propose new features for OWL QL. They are the DL Lite designers.
>
> I tried to give the benefit of the doubt "against" the profiles  
> document. It's still overwhelming in support. I would tend to label  
> Stephens' as fuzzy.
> 	+7.4 or +6.4 (if you don't count fuzzy).
> 	+ 5.5 if you exclude fuzzy and TQ.
> 	-2  or -3 (if you count TQ as negative, which they weren't :))
>
> Given all this, I think that it would be very difficult to de-Rec  
> profiles. It would certainly call for some sort of last call (but  
> that's weird...last call of...a de-reccing?).
>
> I'd suggest that the AC Reps that you are in discussion with send  
> mail directly to the WG. At this point, we've pretty much said that  
> Profiles is rec track and so it'd be a pretty big deal to back down  
> on that. We'd need some heavy ammunition.
>
> So, I think I would not support that absent heavy ammunition  
> regardless of the overall merits unless the situation changed quite  
> a bit. (I.e., it might work to have a delayed CR if the group were  
> extended or something. I don't know.)
>
> Other research or corrections to the codings above would be welcome.
>
> 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