Re: LANG: compliance levels

I'm uncomfortable that "Level 1 RDF Schema on Steroids"
focuses on global domain/range restrictions rather than
local restrictions.  I've found general consensus that local
restrictions are preferable.

I think we want as seamless a progression as
reasonable/possible from Level 1 to Level 2 (so users can
graduate as they appreciate and/or need additional
features).  If we only support global restrictions in Level
1, I'm tempted to suggest that we drop local restrictions
from Level 2.

I think cardinality (preferably arbitrary, but at least
0+/0-1/1/1+/n) needs to be addressed in Level 1.  Most current
DAML+OIL ontologies include cardinality restrictions.  XML
Schemas typically include cardinality restrictions.

I'm torn on the functional property issue.  While I think a
global restriction could be a useful hint to reasoners, I
think this would often be misused.  For instance, a
usSocialSecurityNumber property uniquely identifies a Person
but not a BankAccount.  As suggested by others in the past,
I'd prefer a local property like

  :usSocialSecurityNumber owl:uniquelyIdentifies :Person

Generalizing and going out on limb a bit, I'm concerned that
we're giving RDF too much sway (ignoring charter issues).
If WebOnt is successful, I expect most folks will use it
rather than RDF.  This is typical in layered systems
(compare the amount of application code written to use
10baseT, Ethernet datagrams, IP, TCP/UDP, and HTTP which are
(roughly) successive layers in the ISO OSI Reference Model).
I'm a bit concerned that we're making decisions that will
inconvenience millions of future WebOnt users for the sake
of hundreds of current RDF users.


Received on Wednesday, 24 April 2002 20:26:41 UTC