Lists Home |
Date Index |
On Thursday, Aug 14, 2003, at 20:17 Europe/Berlin, Bullard, Claude L
> I realize that but may not be getting exactly what you
> are saying.
perhaps it would help, if i express my concern in terms which are
foremost in your mind at the moment.
i suggest that, when you decide to warrant a product, you would do so
on the basis of the name of the programming language in which it is
rather, you would likely consider the performance and implementation of
the specific product. you would also likely consider your knowledge of
the supplier, their performance and claims. perhaps even their
supplier. you would also likely consider the intended use. would you
thus, i would expect at least that at least this degree of
expressiveness be available to the descriptions of entries in your ip
just as the name of the programming language is not sufficient, neither
is the name for the set of universal names, nor is the name of a schema
sufficient. it is necessary to name a concrete, rather than a generic,
just as you would need to describe your supplier chain, it is necessary
to describe the context into which the intended process will be
just as you need to describe the intended use, you need to describe the
intended or permitted constituents.
this search space is not flatland and the name of the namespace of the
gi of the root element will not get you there.
> Part of the problem is precisely, which
> combinations are meaningful and I would expect that
> to be down to the level of elements at least, although
> practically, we usually deal with namespaces as
> larger collections, typically, whole languages. Yes,
> versions have to be managed. That is an issue for the
> standard. How do they handle versions now? Different
> means from namespaces to version attributes, and so on.
> Again, I'm not trying to work the problem of natural
> language symbol grounding, only artificial languages
> with bounded constraints, not open ended at the extremes.