[
Lists Home |
Date Index |
Thread Index
]
Sort of. Logical distinctions of names and types
are not always applicable to natural languages if a
context of use by a particular user requires them
to either use very explicit rules for disambiguation
or to accept ambiguity as a norm where the ambiguity
has no real costs. In the *real* world, we accept
and enjoy ambiguity. QNames aren't real world; they
are systemic.
It is dicey to say the name is the thing. Even
the QName is suspect unless one is wedded to the
system identifier, (the URI) and then only as reliable
as it is authoritative. The result of that
will be a struggle to establish and maintain authority.
The registries quickly go from being libraries to
language courts for arbitrating who's zoomin' who.
The semantic web won't be a lot better than natural
language at this.
Fixed simple types are one way to say "in this system,
you better mean this". Fixed values in DTDs are
the same thing. Any data dictionary is basically
just a contexualizing system and if one can have
any doubts, the cost of doubt is borne by the
user. It will be fun to find out if the W3C has
the cahones to accept the notion of pluggable types
rather than simply fixing a subset of XML Schema.
Not new news but it's fun to observe that
commedians are often more succinct about this than
our experts.
My daughter's answer to "why do we drive on the
parkway and park in the driveway?" was "Because
the garage is full, Daddy." Literalist. No fun. ;-)
len
From: Eric Bohlman [mailto:ebohlman@earthlink.net]
>Why do drive up ATM buttons have Braille on them? *
The classic "free feature" scenario; if product B (in this case, keytops with Braille) can do
everything that product A (plain keytops) do and cost no more to make, then there's no point in
producing both, and lots of reason (in this case, the fact that injection molds are expensive) not
to.
How do we decide, when preparing a spec, whether a feature is a "free feature" as opposed to a
"creeping feature"? Presumably not the way the XSD WG did :)
>Why does Hawaii have Interstates? *
Premature name binding. The program's name was wedded to its original scope, which later expanded.
This one reminds me of "wParam" in MS Windows programming.
>Why do pants come in pairs but bras come in singles? *
The name of the interface was too closely tied to the implementation, which changed after the name
stuck (also reminds me of wParam).
>Logic won't always get us all the meaning.
>
>(* thanks to George Carlin for making us think about thinking)
Did you consciously intend to make all three examples relevant to the issue of whether or not a
schema language should specify a fixed set of simple types?
|