Re: [RT] rethinking the need for inferencing

From: Ben Hyde <bhyde_at_pobox.com>
Date: Tue, 7 Mar 2006 08:50:54 -0500

Some added random thoughts.

What you desire in the RDF chewing tool chain depends a lot on what
your aspirations are. There is a tension between getting-things-done
and keeping-options-open. That tension is very similar to the
tension between and closed world and open world.

I find RDF is sweetest when it is empowering me quickly inhale a lot
of diverse data from an open (messy) world. I can defer resolving
the conflicts between the collected data and get the quick
satisfaction of obtaining a pile-o-data. Later I can get to work on
puzzling out how to make this data play together. So another element
of this tension is between lazy and greedy approaches; or as I prefer
to think of it between fun and dutiful.

To summarize:

getting-things-done v.s. keeping-options-open.
closed-world v.s. open-world.
greedy v.s. lazy.
dutiful v.s. fun.
neat v.s. messy.

Part of what I perceive is happening as we in Simile start to covet
some tools like the ones that Stefano outlines is a rising desire to
get things done. That demands zippy algorithms so our users just
can't stop playing with our toys. Only the most committed will play
with a toy who's feedback arrives 10 seconds after you pull it's string.

In the final analysis that's why we covet rewrites like mapping
a:maker and b:author into a single broswer:creator. This tension is
one that I suspect arises in much RDF work. You begin in the open
world but you transition into a closed world where things can get
done in volume.

These seems perfectly natural to me. The world is open, but the
individual actors in that world aren't. They each have a close world
model and interfaces where they normalize so they can deal.

Anytime the actors do this they are going to lose data that could
turn out to be extremely critical. If a:maker is populated with the
URI's of machines and b:author is populated with humans mixing them
together loses something.

To my mind putting tools like the ones Stefano is describing into the
hands of RDF worker bees licenses them to make lossy generalization
like that one.

Such tools don't line up with my little table of tensions above.
Tools like this are lazy and fun, but they also greedy and let you
get things done. Having a powerful infrencing engine in your tool
kit doesn't aline up with the tensions above either.

  - ben
Received on Tue Mar 07 2006 - 13:49:31 EST