implementation report issues for moving through CR

WOWG-
  I am resending the message I sent earlier today with the subject 
re:imports  (archived as message 0315 in February).  This is just to 
get the appropriate header and start what I believe is an important 
thread for our continued progress to move through the recommendation 
track process

================= resend of main content of 0315 =====================

Our hope is to move from Last Call directly to Proposed 
Recommendation (i.e. skipping the Candidate Recommendation phase). 
To do this, we have to show  TWO EXISTING IMPLEMENTATIONS of every 
feature in the language.  Proving something implementable is not 
enough.   The mail from Sean and Jeremy, for example, shows two 
implementations of Imports, and thus that may be sufficient.  We also 
need same for everything else in our design.  Our language won't be 
stable, however, until after LC.  As a result, if we don't want to 
have to have a long CR period, we NEED people to start implementing 
NOW (despite the instability) and then to tune as the language 
finishes (at this point we are not making major changes likely to 
require significant implementation changes)

I've started working on our implementation report, and to move out of 
LC I think we need the following:
  i. Another independent implementation of Owl Lite.  The Univ of 
Maryland will do one, but we need another.  (Note: I do not believe 
it counts to say that all DL implementations also implement Lite, 
because that doesn't help us validate the decision to have Lite as a 
separate sublanguage).
  ii. Two separate DL implementations that have actually been shown to 
pass all, or at least most, of our DL tests (Euler will be needed in 
the Full section).  While I don't doubt there are a number of systems 
around that COULD pass our tests, someone needs to actually show they 
work.  I'm hoping the "Manchester connection" (I.e. University and/or 
Network Infernece) will do one - someone needs to volunteer to do 
another (this could be as simple as writing a tool to cooerce our 
test cases into Racer or other such system)
  iii. As far as I can tell, none of our documents have been changed 
to address the issue of what is expected in datatypes.  This means 
that as it currently stands, we need to produce two sound and 
complete implementations that include all the possible rules for all 
the xsd: datatypes and their combination.  My hope is we'll fix this 
by removing the requirement for sound and complete datatype reasoning 
and put in something rational (OIL is a good model), but if not, we 
will need these two implementations
  iv. I think we will need a second participant to do a Owl type 
checker like Sean is doing -- this is because we make a big deal 
about this in the conformance part of Test.  (If we were to water our 
wording down a bit, we might be able to get away with one, so I'm not 
too worried  about this one)

Here's the good news
  we have well over 100 ontologies that cleanly map from DAML to OWL, 
so we have lots of examples  (currently all the ontologies in the 
DAML library which can validate against DAML can be mapped to Owl 
using the UM converter)
  we have a number of DAML tools that are being adopted for OWL
  we have several implementations of Full being done (I consider Euler 
one of these, cwm another, and we have a student looking at mapping 
Full to a FOL or HOL prover)
  we have a couple of validators coming along - Mike's and Sean'si
  we have at least two parsers (Jena as is, and a new UMd one) which 
can create correct triples for OWL (Jeremy, I think this is true 
based on my understanding of what is in Jena - I know you'll do more 
eventually to make it more OWL aware, but I think it already counts 
as a parser - if I'm wrong, please help me out)
  Within the next week or so we will have at least one web site that 
is entirely powered by RDF/OWL tools - it will demonstrate the 
interoperability of a number of the pieces above.

  So - we either need to figure out that we can do all the things in 
the "to be done" section by end of LC period, or we need to have a 
Candidate Recommendation period with a call for implementations. 
Given the fact that the number of working things (the good news) 
outweighs the still needed part, I'm still hoping we can skip CR.


-- 
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	  240-731-3822 (Cell)
http://www.cs.umd.edu/users/hendler

Received on Thursday, 20 February 2003 11:40:29 UTC