Lists Home |
Date Index |
Michael Champion wrote:
> Not to channel Udell too often in one day, but how is that different
> from HTML with metadata inserted as a side effect of CSS styling?
> Anyway, I'm definitely not disagreeing, this is a very useful approach.
I guess in two ways. One is structural - you could end up laying out
@class tags in ways that's it hard to pull back structured info (be
it 3-tuples or somethin). The other is idiom - that is, it's not
idiomatic to use URLs in @class attributes (I don't recall if CSS
syntax would support such URLs).
On the other hand, doing something as simple as hacking blogging
catgeories (many people have maybe half a dozen or so) onto URLs can
give you quite useful metadata. For example, I tend to use WikiWord
syntax on the basis that someday I'll get of my arse and merge that
namespace with my Wiki. So you can prime things in low effort ways.
In practical terms @class is still very useful; I have a colleague
in Propylon, Clyde Hatter, who leverages @class attributes in markup
to great effect.
> So "metadata" is "content" transformed into a more conveniently
> machine-processable form?
I think so; it can be good enough or useful enough given the burden
that's been taken off the user.
> OK, that does explain a lot of points that I
> wasn't understanding in your, or Dare's or Joshua's arguments. It's not
> what I thought of as metadata, or what Doctorow is flaming. I guess we
> need a meta-model of metadata to keep this all straight!
:) Fwiw, I don't buy into the more GOFAI visions. I don't see that
we'd neccessarily reach a semantic inflection point without a
statistical/signal processing layer in the architecture doing the
bulk of the work.