Re: PR/CR Page

At 12:05 AM +0100 7/22/03, Charles White wrote:
>At the bottom of the PR/CR page, the following showed up recently.
>
>Candidate Recommendation Exit Criteria
>
>integrate any changes to RDF Core specs
>2 complete OWL Lite consistency checkers (i.e. 2 which pass all OWL 
>Lite consistency and inconsistency tests and moreover claim logical 
>completeness)
>at least one reasoner passes every test that is not an extra credit test
>two reasoners implementing (different) substantial subsets of OWL DL
>two reasoners implementing useful subsets of OWL Full
>two owl syntax checkers passing all tests
>
>Did we ever agree to this? I remember seeing this in a message from 
>Jeremy Carroll, et al, but I don't think it was ever agreed upon by 
>the group. Or did I miss something? I looked in the minutes and 
>didn't see any evidence of it. I thought we had agreed that it was 
>not necessary (although possibly desirable) to have all these tools 
>built before recommendation.
>
>chas


Charles - the issue of what the exit criteria are is one of the 
things to be discussed on Thursday.  Jeremy is the only one who has 
proposed explicit criteria, and those are what we based this on.  It 
is worth noting that Dan and I went through these and believe that it 
will be  possible to complete these in 4-6 weeks, so we think they 
are reasonable.  Here is why I (note I say I and not we) think that:

1 - 2 complete OWL Lite consistency checkers - this is not that 
difficult to do, the algorithms for this subset of DL are relatively 
well-known and implementations don't seem to be that hard.  The 
University of Maryland PELLET prover and NI's Cerebra will likely 
meet this need in the near future.
2 - one reasoner that passes every test that is not extra credit - 
this one is the one I'm personally most willing to take out - in part 
because the definition of passes every test is a bit slippery -- if 
we accept Jos' definitions, than Euler is quite close to meeting this 
criterion - if we don't, then I would argue we might rephrase this to 
passes every entailment test (axiomatic reasoners cannot necessarily 
pass the consistency tests)
3 - two reasoners implementing (different) substantial subsets of OWL 
DL: I believe we already have these between FACT, Cerebra, and 
VAMPIRE.
4 - two reasoners implementing useful subsets of OWL Full - I think 
we have this in Euler and cwm/Otter if the latter is made to attempt 
our tests.  If not, several other axiomatic systems are under 
development that are likely to pass many of the Full tests.
5 - two owl syntax checkers passing all tests - I believe Sean has 
completed one of these, and I expect my group to have one by end of 
August - I think there are other groups, including the Jena group, 
working on this as well.

So, while I'd be willing to see these slightly loosened (for example, 
I might prefer "a large majority of tests" instead of "every test"), 
I think we could actually meet these criteria by the end of the 
proposed CR (dates to be established, but Dan and I believe mid-Sept 
should be our target - more on that in a later message)

WG - this would be a good thing to discuss before Thurs on the 
mailing list -- Chas, do you have a suggestion for an alternative?

  -JH

-- 
Professor James Hendler				  hendler@cs.umd.edu
Director, Semantic Web and Agent Technologies	  301-405-2696
Maryland Information and Network Dynamics Lab.	  301-405-6707 (Fax)
Univ of Maryland, College Park, MD 20742	  *** 240-277-3388 (Cell)
http://www.cs.umd.edu/users/hendler      *** NOTE CHANGED CELL NUMBER ***

Received on Monday, 21 July 2003 19:19:25 UTC