Re: Well Behaved RDF - Taming Blank Nodes, etc.

>
> What complete nonsense.


Ah, semantic web. Other forums would be content to debate pin-head
angel-count, but here we get cheerfully stuck arguing (dismissively!) about
whether we mean a *specific* pin! No wonder SPARQL didn't initially have
COUNT.


Here's the one thing of which I am very sure: This forum/community/whatever
suffers terribly from not distinguishing between (at least) two very
different activities that happen to both involve data-modeling in some
sense.

One of them is concerned with modeling, querying, publishing, maintaining
and otherwise messing about with what we have traditionally thought of as
databases, or the kind of data that we have traditionally put in databases.
That is, things you *could* do in relational databases, except maybe
relational databases are not the end of human progress in this domain, and
maybe something under this conceptual heading of data-graphs could be
better.

This world is no more fazed by the burden of inventing URIs than it was by
the burden of inventing primary keys for SQL tables. It doesn't suffer any
existential doubt over whether the picture of a room inside a building is a
picture of the building. Sure, in one human sense it is, in another it
isn't. But both senses can be modeled, so we pick one and move on. On a
real-estate site, pictures of the rooms of a house for sale are clearly
pictures "of the house". In some other database, for some other purpose, we
may make a different decision. And we understand that externalizing
literals is always more flexible and forward-looking than not, but
sometimes we optimize for the present. The decisions are practical, because
the problems are practical.

The other world I'm hesitant to characterize, because I clearly do not
really live in it. But I've visited it. I've taken pictures of it. Or maybe
they were rooms inside of it, I guess I'm not totally sure. But I've eaten
the street food there. It's yummy, and when I come home I'm sad that nobody
where I live cooks it correctly.

This other world I love visiting is concerned with how we say things that
can be said. In this world, it's a real and important question whether we
mean a particular pin or not, or how we say one thing or the other. We're
not making a database of pins for a pin-seller's web store, we're talking
about the nature of truth. And clearly the nature of truth allows for open
worlds and existential quantifiers and chainable inference and transitive
closure and whatever.


Both these worlds have, historically, adopted toolsets and compromises that
serve their purposes unevenly. The SQL-centric instantiation of the
relational idea has all sorts of well-established limitations, as does the
current state of RDF/SPARQL/OWL/etc., and we split our time between living
with these limitations and imagining ways to not have to live with them.
This tension between application and theory is productive. It's
definitively productive, I think, or at least I don't know any better way
to make progress.


But I'm not sure the tension that comes from conflating the different *
worlds* is productive. At least, I've personally been watching it happen
for 7 or 8 years now, and it doesn't seem like I'm appreciably closer to
being able to use SW tools for my own data problems. And yet when I visit
the other world and come back, the nominal other-world street food here is
still not quite right.

So while I'm not sure what the solution is, and probably somebody with the
luxury of visiting the semantic web as a tourist isn't the one to write the
zoning laws, I will politely suggest, in the margins of the customs form as
I leave with my souvenirs yet again, that the community might benefit from
distinguishing between these purposes in its discussions, and not applying
objections that obtain in one world to ideas designed to operate in the
other.

g