Re: comprehension

Jeremy Carroll wrote:

>
> Given that Pat, Peter and Jonathan all think that my proposal is fatally
> flawed because of the difficulty of doing comprehension, I had better
> address this issue.
>

To be clear, nowhere have I stated that the "comprehension" proposal is
fatally flawed. Indeed I stated that it looks "promising", but certainly you
must agree that it is not as developed as for example Peter's description
(sic) of how dark and light triples correlate with standard description
logic. Similarly Pat makes a fairly standard and well known argument about
the difficulties inherent in certain ways of extending a language using the
language itself. Both of these arguments appear to me to be based on well
known, and well studied fields in which many many papers have been written
and there has been a great deal of experience. So yes, it is not at all
clear to me that anyone has ever actually done what you propose and gotten
it to actually work on actual data in a robust fashion. If I am wrong please
explicitly cite the software and application etc. Unless that is the case,
then yes it is a research project, and yes research can be extremely
promising. Just so.

Twisting my words, and arguments, in such a fashion does little to give me
confidence in your other words and arguments.

My concern is, and has been, that it will be difficult to consensus on this
in the very near future -- for example -- are you _certain_ that your
proposal as outlined is _complete_ and nothing will need to be changed _at
all_ given feedback from members of the WG? My strong suspicion is that the
nitty gritty details of your proposal still need to be worked out, and at
present I don't have a crystal clear idea of the subtle benefits and
problems that might occur. I did say it was "promising", but in many fields
it is attention to detail that makes the difference between inadequate,
barely adequate and great.

I will present a risk analysis as I see it, of the various proposals and
solutions, but for the moment, statements like:

> Moreover, even if these rules are mistaken, implementors will assume these
> rules unless contradictions are explicitly pointed out to them. Thus any
> paradox found in the future contained in these rules is an issue for owl
> that will require clarification. This need is independent of whether or
not
> we endorse these rules.

greatly concern me.

Jonathan

Received on Friday, 3 May 2002 11:49:51 UTC