RE: namespace usage as assertions

> every namespace is a scope which contains every possible legal XML name

This is the level at which I have trouble with the assertion that "a
namespace is a vocabulary"  -- I'm having a bit of trouble wrapping my mind
around a definition of "vocabulary" which has an infinite set of members.

But maybe that's not as unreasonable as it feels.. English often seems to
have a similarly infinite set; you can't reliably say that "Ximplacic"  or
even "tkpbldqnxtrz" aren't words, only that you aren't familiar with them.
New words are coined continuously, and the only real test of whether they
are part of the language is whether they're widely adopted and understood.

(Even a recognized authority is probably insufficient. Did anyone in France
actually accept the ruling that CDs were to be referred to as "disque
audio-numerique" or something like that?)


Note that the same word may be understood very differently in different
contexts, and that those meanings may over time as new contexts arise and
old ones fall out of use. Analogy, simile, metaphor, homonym, homophone,
simple overloading... These may not be "best practice" but they're common
and powerful.

Maybe namespaces should be viewed less as vocabulary per se and more as
markers of context. After all, their essential purpose was to disambiguate
cases where the same "word" (name) might be used in several different ways
in the same document, and where grammar alone was insufficient to
distinguish them... and to provide a recognizable name for the context, so
that one can discuss which of these cases the word resolved to.

If so... Maybe a namespace is most equivalent to a jargon. At that level,
maybe it does imply some semantics. Unfortunately we have no architecture
for expressing those semantics at this time, but maybe that can be added in
the future.

This does tend to reinforce my belief that relative namespace declarations
are bogus, though. I do _not_ want that context changing when my words/data
are taken out of their original framework ("This movie was a great
disappointment; don't see it!" -> "... great .... see it!").


A vocabularly definitely _isn't_ a language. To make it one you need to
associate it with syntactic and semantic rules. There are creole languages
which use primarily English vocabularies with  significantly non-English
syntax and semantics.


I do see where folks want to take Namespaces, as a building block toward
more complete concepts of language. But namespaces by themselves simply do
not carry that load, probably can not, and should not be expected to. It is
sufficient if they're good at that part of the problem they _do_ address.

Which still brings me back to "Forbid" as my first choice. If we
absolutize, we're going to pun ourselves to death. "Send the bellboy at 8AM
to knock me up" may malfunction very badly indeed if your vocabulary is
indicated by a relative reference to "English".

______________________________________
Joe Kesselman  / IBM Research

Received on Friday, 23 June 2000 10:49:16 UTC