Re: URIs quack like a duck

This discussion is now about the logistics (and the requirements) for implementing a
referential integrity mechanism. XML 1.0 is silent on this question. 'Namespaces in
XML' seems aware of the difficulties, but punts by mandating character-by-character
comparison:  introducing a fixed intermediate point of reference specifically
abdicates the job of solving this variable-upon-variable problem. For inherent,
characteristic reasons of their own the 'higher' XML specifications--schemas, query,
link resolution, and the RDF family--not only will have to face up to, and solve, the
problem, but each will have to do it in a way which is consonant with its own
premises and therefore unlikely to be perfectly applicable to the peculiar needs of
the others. This disparity between the referential needs of RDF and the (lack of)
mechanism provided by 'Namespaces' has already animated most of the discussion on
this list.

Implementations in running code of different specifications of the XML family will of
course understand, and apply, the language of any standard in light of their own
needs and biases. This should be recognized and accepted. Any specification for a
referential integrity mechanism (or perhaps the different referential integrity
mechanisms, considered as a whole across the XML family) should define a range of
algorithms, and the syntactic structures which support them, leaving it to the
individual specifications--and perhaps in some cases to the particular applications
which implement those specifications--to use the simplest and most efficient approach
which answers their unique needs. Approaching the problem in this way does not
undermine the goals of standardization. In fact, to recognize that there is a wide
range of needs for referential mechanisms and to specify different, but often
cumulative, solutions across that range is to take up the real work of
standardization. Instead of the Procrustean bed, which will eventually be shunned by
needs of every size, we have the opportunity to promulgate well-coordinated, modular
solutions which combine into well-tailored pipelines of processing for the
circumstances at hand.

Respectfully,

Walter Perry

John Cowan wrote:

> The difference is precisely in the use of relative URI references like "foo/bar".
> In my ("data scheme", "new synthesis") proposal, every reference to
> the namespace name "foo/bar" is a reference to the same namespace,
> independent of context, because the name is mapped to the URI
> "data:,foo/bar".  Accessing "foo/bar", on the other hand, produces
> an entity body that is emphatically dependent on context.

Received on Thursday, 1 June 2000 13:46:33 UTC