Lists Home |
Date Index |
On Tuesday, Oct 21, 2003, at 10:49 Europe/Berlin, Michael Kay wrote:
>> On Tue, 2003-10-21 at 09:45, Joern Clausen wrote:
>> I repeat my question, which was
>> not answered yet: Why are only elements and attributes
>> subject to a namespace definition, but not entities? Has this
>> been discussed while preparing the Namespaces REC, and was it
>> decided not to go this way for any good reason? Pointers to
>> mail archives are welcome.
> Layering. There is already far too much interaction between the
> and logical layers of XML, e.g. the rule that tags in an external
> must be well-balanced. It would ideally be possible to do all the
> lexical processing without any interaction with the syntax processing.
are we talking about the same xml which has over two dozen lexical
contexts? the recognition of which depends very much on syntax
> The fact that namespaces are not even part of the XML syntax, but a
> semantic layer on top, would compound the problem even further.
> <a xmlns:foo="&foo:entity"/>
it all depends on your point of view. as noted in another message there
are views of processing, like
> in C++ [in which]
> variables, class names, and functions can be namespace-qualified, but
> preprocessor macros cannot.
and then there is a view, that if names are treated uniformly as
first-class data, one does not create a situation, where
> About half the code I write for XML processing deals with
> namespaces. I'm not looking forward to making that 75%.
in fact, from a technical standpoint, as soon as one admits that an xml
processor is not a text processor, it is not hard at all to treat all
names uniformly. qualified entity names are localized to productions
67-72 and (probably not only in cl-xml) pass through a single
resolution operation. that is, complexity cannot be the reason.
when the question was posed, years ago, the answer was, that one had
notations, which would provide any necessary qualification, so
explicitly attaching a uri to the entity name (or, in the case of the
age-old thread, a pi target) was not necessary. i was not the only one
curious - there are threads in xml-names-issues on the topic.
upon skimming through those messages, this reader is struck by how, on
one hand, so little of the purported semantic benefit has played out,
while, on the other hand, no one wants to do the one thing would be
easy - that is, to really treat qualified names as a uniform means to
encode unambiguous names.