[
Lists Home |
Date Index |
Thread Index
]
Simon St.Laurent wrote:
> 1) What's that URI really mean? Don't ask, or we'll never get home. In
> fact, it's probably better not to ask _anything_ about URIs, or we'll
> all end up on www-tag[1], with its 288 mostly URI-permathread messages
> this month alone.
And for the first time in years maybe, a URI/Semweb thread really
worth following. The difference this time being that the TAG has had
to debate the semweb with an AI legend whose been through all this
before:
[[[
You seem to be operating under a kind of broad-lapels Yorkshire kind
of confidence in the practical: we just need to get down to brass
tacks and it will all be clear, none of this silly academic
shillyshallying about over bloody theorems and the like. But it
won't. Honestly, now, his is where we AI folk, for all our faults,
really do have some practical experience. You DONT get operating
ontology software just by writing down the obvious practical stuff
that everyone knows, like dates and total amounts. First, people
don't agree about what things like dates really are (is a month a
kind of duration? Think carefully before you answer). Second, even
when they more or less do, after a *great* deal of discussion,
deciding how best to formalize it is still an art, and there are no
ground rules or even broad consensus for many - most - topics.
Things as simple as how to describe a river flowing to the sea, or a
tropical storm, are still research issues. And third, there are no
agreed-on upper level conceptual frameworks to fit all this stuff
into; existing deployed ontology standards do not agree. In many
ways, the 'simple' things that everyone knows are the hardest to
formalize, sine we don't normally have to even articulate them.
]]]
There's probably a dozen more quotables where that came from. Pat
Hayes is forcing the issue to a conclusion.
> 3) While most tools now understand namespaces and namespace
> declarations, many programmers still use local names even in very
> namespace-specific contexts. For one very small example of this, see
> [2], but I've encountered this practice constantly. I think that XSLT
> is the only programming space where this practice is rare, and that's
> because the tool doesn't work with local names alone. (Figuring that
> out causes beginners a lot of pain, but once they're past it, they're
> likely better XML practitioners.)
I'd say that's down to the way XPath models QNames/Elements, which
givent the alternative (prepend localNames with URIs not prefixes)
is neat enough.
> 4) QNames in content seems to be an ever-expanding mess, with no means
> in sight for a normalization method that makes them context-independent.
> (That's largely because there's no simple mapping between QNames and
> URIs - hashes vs. slashes keeps that complicated.)
A mother of a bug.
> To me, the bad factors outweigh the good, but I too have "survived
> namespaces" and implemented support for them in the code I've built.
> (Perversely, I'm even considering extending namespace support to entity
> names in a current project.) That said, I don't look on "survived" as a
> badge of pride - it's more that I've come to terms with toxic sludge in
> our collective basement which we can't afford to clean up.
+28 (days later).
;-)
Bill de hÓra
|
