Lists Home |
Date Index |
Arjun Ray scripsit:
> Arguably this is inherent in the process of fixing names for things, just
> as non-terminals are context-free in practical parser generators and BNFs.
Just so, but although DTD element declarations look like BNF definitions,
they are not -- precisely because an element declaration demands that
the content model it defines be wrapped in an element that is visible
in the instance syntax. (RELAX NG definitions truly are definitions, OTOH.)
> IOW, analytic composition is the principal characteristic of element types
> in this formalism. If you need a different model, give the summation a
> different name. But this is ultimately only because computers don't deal
> with imprecision and ambiguity quite as well as humans do.
The kind we need here is the kind that computers deal with every time a
compiler or interpreter is run.
> The formalism
> does call for a certain amount of circumstantial invention of distinct
> names where humans might have found the same names more congenial. But
> insofar as incidence of the same name does not necessarily mandate the
> same processing (which can be contextually controlled), I'd say that the
> cost of such a naming discipline isn't onerous.
I was rather referring to the inability to give the same name to distinct
things: for example, a title borne by a human being and a title borne by
a book have some similarities, but enough differences that they probably
wind up with different GIs in valid XML or SGML. Backed by a stronger
validation formalism, though, they can have the same GI.
The practice of the XML community has been overwhelmingly to grant the
GI fundamental importance: fulminating against this sociological fact
will not make it go away.
> | The treatment of attributes is also weak,
> In what way? I suppose I'm also asking for an answer to the issue of
> "ontology" I raised earlier: what are attributes for?
Quot homines, tot sententiae ("as many opinions as persons"). There simply
does not exist any generally accepted view of when attributes should be
used rather than child elements. Therefore, it is important for a
neutrally usable schema language to support them as identically as possible,
excepting the obvious (attributes are unordered and can't contain elements
or other attributes).
John Cowan <email@example.com> http://www.reutershealth.com
I amar prestar aen, han mathon ne nen, http://www.ccil.org/~cowan
han mathon ne chae, a han noston ne 'wilith. --Galadriel, _LOTR:FOTR_