Re: Joint Committee telecon tomorrow 17 September

From: Jeff Heflin ([email protected])
Date: 09/19/02

  • Next message: Richard Fikes: "Comments on DQL from Joe Rockmore" 
    "Peter F. Patel-Schneider" wrote:
> 
> From: "Lassila Ora (NRC/Boston)" <[email protected]>
> Subject: Re: Joint Committee telecon tomorrow 17 September
> Date: Wed, 18 Sep 2002 11:02:59 -0400
> 
> > Related to our discussion yesterday about including/importing: I wanted to
> > stress the fact that although things like complete reasoning are different
> > from things like whether some file can be accessed at some particular time
> > on the web, we (the Joint Committee) should be designing our things
> > (representation languages, query languages, rule languages, etc.) in such a
> > manner that the nasty eventualities of the web can be dealt with.
> >
> > If we do not take the architecture and the realities of the World Wide Web
> > into account, we are not doing anything really new. Then our work becomes
> > "just" KR. And furthermore, I believe that in order for us to be successful
> > in furthering the Semantic Web, dealing with these details is a strong
> > requirement.
> >
> > I am not suggesting that we fully immerse ourselves in the problems and
> > issues of distributed, asynchronous systems and communication protocols, but
> > I am suggesting that whatever we do has to make life with distributed
> > processing easier. Given that our stuff gets built and lives on top of many
> > earlier web standards, we can take full advantage of these if we expose some
> > of their features (HTTP caching rules for example come to mind).
> >
> > I'll get off the soapbox now...
> >
> >     - Ora
> 
> Well, my viewpoint is that it is sufficient to
> 1/ not require committing to information from other ontologies unless an
>    explicit directive so stating is present
> 2/ behave reasonably when referenced documents (or other resources, such as
>    computing resources) cannot be found, or accessed, or are unresponsive,
>    probably by continuing reasoning but not labelling results as complete.
> 
> The first is a matter of the logic, the second has more to do with
> interface design.
> 
> Peter F. Patel-Schneider

Wow, I actually think I agree with Peter on something about imports. ;-)

    This archive was generated by hypermail 2.1.4 : 09/19/02 EDT