RE: universal languages

On Sat, 3 Feb 2001, Miles Sabin wrote:

> Sandro Hawke wrote,
> > One nit, in the name of clarity here: you only need pairs for
> > directed graphs (DGs).  Triples give us directed labeled graphs
> > (DLGs).
> >
> > I think you can represent a DLG with a DG, so maybe it doesn't
> > matter.
>
> Indeed you can.
>
> > But it does raise the question: we're using n-tuples to
> > represent information.  Is there any clear reason to use n=3 as
> > our fundamental structuring?   I think it's more conventient
> > than n=2 and makes self-reference easier than n=anything (which
> > most logics seem to use).   But is there a solid theory behind
> > that?
>
> Given that arbitrary n-tuples are reducible to pairs I don't
> really see how there could be.
>
> Scholars of ancient and medieval logics will be very familiar
> with the idea of taking subject-predicate-object as a fundamental
> form ... some people used to have quite spooky ideas about the
> significance of the number three in those days too. Whether it's
> really appropriate to take SPO as foundational for a post-Frege,
> 21st century enterprise, is another question ;-)

While it's easy to get carried away regarding the mystical significance of
3-tuples for the Web and for RDF 1.0, I believe there's a useability story
as to why 3-tuples (nodes and arcs) are a good basis for a mainstream
metadata system. People 'get it'. They might not get the translation into
XML syntax, but the basic observations that underpin RDF 1.0 and the
Web itself should (after we fix a few nits) play well with whatever
languages we layer on top.

The Web is about the relationships between a world of name-able thing; we
call  them 'resources', and use Web identifiers (URIs) to name them.
Those things fall into named categories, and stand in named relationships
to one another. This is how the Web has always worked, and how the Web
will always work. We're just beginning to make explicit some details that
haven't mattered much to Web applications until XML and RDF arrived. RDF
1.0 is a start down that road. One of the biggest confusions for RDF 1.0
implementors though was the nature of the URI/Resource relationship. These
concepts underpin a huge number of W3C specs, and it's time to see those
problems fixed / clarified across the board. Griping about RDF 1.0 being
simplistic misses the point. The Web is simplistic; if this agent
language we're discussing on RDF-Logic is going to work, it's going to
have to work with, for, and on the Web. Content authoring and management
tools will need built-in support for creating and editing expressions in
this language. And that's where the usability issues bite.
Object-related-to-Object is a nice simple conceptual model. You could
teach it to school kids. While there are many reasons for wanting n-ary,
you lose the immediacy that comes with binary relations. I can imagine a
world where Web browsers have bookmarking systems based on the RDF model
(Netscape 6 does just this). And I can imagine mainstream Web users
understanding how such a bookmark system might support the representation
of richer relationships amongst the pages they visit. I have high hopes
that the Mozilla folks will start to explore this. But the notion that we
can make these kinds of app much more complicated than typed links without
losing most of our audience fills me with scepticism...

dan

Received on Saturday, 3 February 2001 15:05:42 UTC