OWL fragments

All-
  I forget sometimes that this new working group is not really a  
continuation of the last one and that it doesn't have the context of  
either that milieu or some of my more public critiques of some  
aspects of OWL (despite being one of the main proponents of OWL  
overall).  So the tone of my last message may have not gone down well  
-- I apologize, it's sort of a carry over from some of the earlier  
comments I made on the public-owl list and some other places, and not  
that constructive for this forum.
  Basically, while I think the work in OWL 1.1 on fragments was a  
good piece of theory, and I've cited it a number of times in papers  
and articles I've written, it was based primarily on theoretical  
issues, not on the implementation issues being faced by Wed  
Application Developers who are starting to get excited by the  
Semantic Web.  I think some of this has led to a backlash that we  
need to overcome - for example one company doing a lot of RDF work,  
and moving its stuff towards distribution via the Apache software  
foundation (i.e.  a big deal) has said publicly it won't include any  
OWL (all for mistaken reasons which I've been in conversation with  
them about) - what it really boils down to is they don't need all of  
OWL right now, and they're moving towards adding some rules stuff, so  
in fact they are reinventing things already in OWL because we didn't  
"package" them right (for want of a better term).
  So, I am hoping that this new Working Group can apply some of the  
lessons learned from the positives and negatives of OWL adoption in  
the 3 years since the spec came out, and can work hard to create some  
"on ramps" to OWL for those who are still just getting involved in RDF 
(S) - "Web 3.0" needs us!
   I didn't meant to insult the quality of anybody's work, nor to  
imply things done to date weren't useful, I just meant to say we need  
to be more careful in this working group to include the needs of  
developers.  For example, in a study of some RDF applications that  
have used OWL, inverseFunctionalProperty on datatypes turned out to  
be used in almost all of them - in our current document, this means  
they are in OWL FULL, which is not untrue, but is definitely somewhat  
misleading.  When we fix some of these disconnects in the new WG, we  
should be sure to backfill and cover these sorts of issues
  Anyway, as punishment for my outspokeness, Alan has asked me to set  
up a Wiki page and lead some discussion of the fragments stuff, and  
to document some of the fragments that I've identified in work with  
Nokia, Oracle and some others.  I will do this as soon as I can
  cheers
  Jim H.

  

Received on Thursday, 11 October 2007 16:53:27 UTC