Re: accuracy check on OWL-DL reasoners

>>
>> Michael Kifer wrote:
>>>  My argument was that this should not stop us from including
>>> things that are a bit challenging and I gave OWL as an example.
>>
>> Agreed.
> 
> So I've no idea of the detailed context for this discussion but it seems
> to me there is quite a lot of difference between a standard where you
> are expecting people to implement or extend reasoners and one where one
> of the requirements is "no new implementation, just translation".

Well, I think the distinction between translation and implementation is
not that clear cut.  For example, if you have a horn rule reasoner
without equality, but you receive a BLD rule set with equality, you can
still translate the rule set to your own horn rules language, because
equality is axiomatizable using horn rules (you just need many of them).

Another point is that if we exclude too many features, the format will
not be very useful to many people.

With the decision we made at the face-to-face to specify a Core subset
of BLD (Datalog without equality) and leave BLD as is (i.e. full horn
with equality) I think there will be many implementations which support
at least one dialect (Core), while still having a dialect which allows
people to do a little more (BLD).


Best, Jos

> 
> Dave

-- 
                         debruijn@inf.unibz.it

Jos de Bruijn,        http://www.debruijn.net/
----------------------------------------------
In heaven all the interesting people are
missing.
  - Friedrich Nietzsche