RE: rdf as a base for other languages

Jos,

I am not sure if this response will make it into the WG list but please
forward it along if not.

>
>
>
> (this is a reply to Jonathan's concerns about RDFCore and I
> thought that it would be good to cc the WG to get more feedback)
>
> Jonathan,
>
> > I think the question is whether RDF is good enough basically as
> is, modulo a
> > few syntactic touch ups and clarifications, or needs to be
> revised perhaps
> > in a substantial fashion.
>
> 'perhaps' is the right word here :-)

I hope you realize that I am somewhat impartial to these arguments. Indeed I
got involved in this long rdf-logic discussion precisely when I _defended_
RDF as less than totally worthless ... the initial position taken by a
surprisingly large segment of the rdf-logic group, and composed of DAML PIs
etc.

You must realize that these individuals have considerable experience in
matters of knowledge representation etc. and I was genuinely quite surprised
to learn of their collective disatisfaction.

Instead of saying that "RDF is worthless", something I find less than
constructure (though indeed may have been successful as a call to wakeup),
we have been able to elucidate some particular problems with particular
areas in RDF ...<strong-emphasis> and it turns out that these areas are
precisely the areas that the WG has identified as problems
</strong-emphasis>

>
> > Correct me if I am wrong, but you seem to be arguing _for_ keeping RDF
> > basically asis, yet are a proponent of N3.
>
> True and let me try to explain why.

Look, I am not sure why people are being so stubborn on these issues ... you
admit there are problems, but rather than fix the problems, insist on a
whitewash over these issues. The DAML folk or at least many of them,
consider these problems serious. What is the resistence to addressing these
concerns using the _best_ method rather than a less than adequate method?

And on record: I am entirely _convinced_ that this can be done with little
damage to current RDF systems and software ... I will be willing to produce
code backing these claims if these proposals are considered seriously.

[explanation snipped]
>
> > I think N3 is terrific,
>
> I more than agree with that and the {} is more shorthand
> than anything else to "identify things by their contents".

yes but the simple _fact_ is that N3's CWM represents statements by a quad
not a triple. Where is "context" in the RDF 1.0 M&S? That is a major point,
the concept of "context" is important enough to be represented in the
statement tuple. To try to do so using triples requires reification and
containers and a single statement becomes 6 statements. What a mess.

Reification in its current form has to go. Regardless of whether it works or
not. The implementation has not had traction. It is a hotspot. Cut it out
and let RDF heal (speaking as a surgeon :-).

>
> > but also think that it changes RDF in a
> > _substantial_ fashion. That's ok, but why don't we just admit
> that change is
> > in order, and get on with the task of creating the best RDF
> that can be had.
> >
> > If you are using N3 as an example of what can be done with RDF that is
> > confusing, rather N3 should be used as an example of how RDF
> should change.
>
> I refer to my above *could*
> and in RDF/xml it *could* be done with rdf:parsetype="log:quote"
> and that is a perfectly evolvable approach.
>

I think that extending the RDF 'model' is also perfectly evolvable. Indeed I
have started to edit RDFFilter to emit the tuple syntax I have proposed.
Turns out that most of the editing is cutting out code needed to handle
containers and reification etc.

Better yet: Let's see an implementation of CWM and N3 using XML syntax and
some current RDF 1.0 compliant software so that we can see how different
abstract syntaxes perform.

Note: I still am a proponent of the _conceptual_ triple
<predicate,subject,object> its just that I see the need to label these
triples with <context> at least and perhaps an couple of other bits.

-Jonathan

Received on Sunday, 3 June 2001 12:58:16 UTC