Re: LANG: need to CLOSE Issue 5.6 Imports as magic syntax

Jim Hendler wrote:

> All, please be careful - no one has said that owl:imports is out of the
> scope of the WG.
> ...
> So no one has said imports is bad or shouldn't be
> there - the question we've been debating is whether people are currently
> comfortable enough to believe the solution on the table is one we won't
> come to regret at some later time.

The point of my message was to sketch a very elementary usage-scenario
(one that I cannot imagine the Semantic Web could do without), and make clear 
that it requires the DAML+OIL-like version of imports, which we understand 
quite well.

If we cannot do this scenario, it will be impossible for one person to extend 
another person's  ontology, and to have these extensions be used by a third 
party; we will have thrown away one of the most applauded features of 
DAML+OIL, namely being able to have statements about a class/property 
distributed over different locations, and use all of them in a single 
application.

I agree with you and Dan that "there are a lot of aspects of linking and 
importing yet to be explored", and indeed "we shouldn't codify things now 
that we will later regret". To me this means that we shouldn't codify things 
now that we don't understand now, but it shouldn't stop us from codifying now 
something simple, which we understand very well, and which is required for 
such an elementary and essential sceario as the one from my message.

So the questions seem to be:

- do you agree that the scenario from my message requires a DAML+OIL-style 
imports (along the lines of what Jeff has been proposing). (If not, show me 
how the scenario can be done without)

- do you agree that the scenario is essential, and that disabling this 
scenario will severely limit the deployment of OWL in an open environment 
like the Web?

- do you agree that we have sufficient understanding of an elementary version 
of import as worded by Jeff and others for it to be included in the current 
version of OWL?

Clearly, this leaves out many issues:
- deriving import-statements,
- subclassing import-statements
- distinguishing varies types of commitment to an ontology,
- allowing partial imports, etc
These must (and indeed will) all be left out of the current design to await 
future versions.

Frank.
   ----

Received on Sunday, 10 November 2002 14:48:05 UTC