[
Lists Home |
Date Index |
Thread Index
]
Rick Jelliffe wrote:
>
> From: "james anderson" <james.anderson@setf.de>
>
> > in order to avoid this problem it is necessary to implement a more
> > complete inference mechanism than that suggested by [1] and implemented
> > by microsoft. where one performs constriant propagation in the type
> > domain [2] one can incorporate the declarations in the document entity
> > and "retrofit" existing dtd's.
>
> If namespace-aware DTDs worked so that an element declared with no
> namespace declaration had an implied declaration of a null namespace,
> what propagation is needed?
>
> So given
[ a description of the problem, which entails the solution [1], which
solution is then inexplicably neglected in favor of a shorthand which
effects local declarations, ]
one could rethink the conclusion and decide to indeed "perform[]
constriant propagation in the type domain". that is, as a content model
declaration predicts the paths which would eventually be used in a valid
document entity to propagate namespace constraints, it can just as well
be used within the type declarations. there are cases where the type
model may be either under or over constrained. some of these can be
resolved with scoping rules analogous to those which apply to the
document entity. [2] those which cannot be resolved can legitimately be
classified as namespace-invalid. note that the same issue exists in
other schema forms. it is not a problem there because the interpretation
rules include rules for namespaces. there is no reason such rules could
not exist for dtds.
...
[1] see http://lists.xml.org/archives/xml-dev/200105/msg00160.html and
the discussion which follows for examples related to this topic.
[2] see
http://homepage.mac.com/james_anderson/XML/code/xquerydatamodel/qnames.html.
this is and excerpt from the documentation from the cl-xml
implementation of such a mechanism.
|
