Re: LANG: compliance levels

Frank,

At 5:27 PM +0200 4/22/2002, Frank van Harmelen wrote:
>A small group met at KR'02 (ter Horst, Patel-Schneider, Horrocks, 
>Welty, McGuinness, van Harmelen), discussing the contents of 
>compliance level 1 for OWL. We solicit reactions from those 
>volunteered for this task. Please do this by immediate response, so 
>that we can report back to the WG next Thursday.
>
>We propose to use for level 1 RDF Schema on Steroids,
>(using the terminology from Frank's Thursday 18 April message)
>with additionally:
>- properties can be declared functional
>- datatypes (details depending on resolution by RDF Core).
>
>The main motivation for this choice is aimed at tool developers:
>this level gives tool developers a useful language to aim at that is 
>significantly smaller than DAML+OIL, while imposing as few 
>restrictions as possible on toolbuilders that want to extend beyond 
>this compliance level. Putting in any additional features (such as 
>universal local range restrictions) into level 1 will make it much 
>harder to go beyond this basic level (for example the interaction 
>with existential restrictions).
>
>Written out in full, this amounts to:
>
>RDF Schema stuff
>     primitiveclass  
>     subClassOf
>     subpropertyof   
>     domain
>     range
>     Property
>     named & unnamed Individual
>
>(In)equality
>     sameClassAs
>     samePropertyAs
>     sameIndividualAs
>     differentIndividualAs
>
>Property characteristics
>     inversOf
>     transitive
>     symmetric
>
>Plus: functionality of properties (= at most one value for a property)
>       (with the usual side condition that this cannot be applied to
>        transitive properties, same side condition as in DAML+OIL)
>plus: datatypes (unclear at this moment what this means precisely,
>       pending on RDF Core decisions.

I like very much the way you grouped functionalities. I think it 
provides the right level of granularity for this discussion on 
compliance level.  However, I am not sure about the new proposal. My 
idea of a level 1 was something like a 'basic frame language', one 
which however is not as 'basic' as RDFS.  Your proposal, as Mike also 
points out, focuses on properties, rather than additional machinery 
for slot specification.  My intuition is that this is not what 
'basic' users of OWL will need.  What I would like to see is the RDF 
Schema stuff + machinery for specifying value/range/cardinality 
restrictions.  We could then discuss whether to add (In)equality & 
Property characteristics, but I guess these extensions will greatly 
complicate the implementation of reasoners for basic OWL, and may 
make pointless to have a distinction between level 1 and 2.

- Enrico

Received on Thursday, 25 April 2002 10:02:55 UTC