namespaces in xml 1.1 requirements (WD-xml-names11-req-20020403)

greetings;

one cannot object to a change which permits binding any given prefix to
the "null namespace". it entails no significant modification to
namespace management algorithms. if anything, it will simplify them, as
it removes a special case.

if the motivation were just that, one could accept it on principle.

this reader is surprised, that the expressed motivation has nothing to
do with decoding per se, but instead describes an attempt to rectify
decisions which were made in the standard for describing documents - the
infoset. or perhaps more concretely in some model implementations which
conform to that standard. the requirements document alludes to problems
which can arise when interpreting reified namespace properties in a
concrete model.

where, alternatively, the reference from the prefix to the namespace
name is folded to a constant at the point where it is parsed, that is to
say, where names are interned, those problems, which can arise because
of artifactual namespace bindings, do not occur. the namespace bindings
act not as an environment for interpreting names within a document
model, but as declarations to appear in the encoded document. when they
are used by the serializer, it is for the convenience of the human
reader. in-scope namespaces can be derived on-the-fly by a serializer
and need have no significance beyond the scope of the encoded document.
the original declared namespaces need have no affect on the canonical
serialized form of the document and could be suppressed.

several examples to this point would be available in the xml-dev
archives. others are also available in the cl-xml distribution, in the
file xml:tests;xparser;test-qnames.lisp

to turn around and change the rules for decoding in order to alleviate
problems which arise as a consequence of an unfortunate concrete model
and/or method for encoding is, well, pervers.

if one is contemplating changing the way names are interpreted, one
could conceivably recognize that it is possible to interpret universal
names in a dtd. while that might stretch requirement 5, it would be well
worth the effort. 

...

Received on Wednesday, 3 April 2002 18:52:46 UTC