[
Lists Home |
Date Index |
Thread Index
]
Dare Obasanjo wrote:
>I'd be more interested in someone who can point out the the difficulty
>caused by namespaces in XML that don't boil down to philosophical or
>aesthetic arguments.
>
Namespaces should point to schemas, being the actual vocabulary
definition pointer. Even better, it should point to an an RDF document
(after all, RDF is about metadata) or perhaps something like an RDDL
doc, or whatever.
Secondly, a namespace URI designed *not* to actually point to something,
should not be a URL or any other type of URI that uses a scheme designed
to be used for resource retreival.
Attributes, the base of numerous debate around XML are also tortured by
the XML names recommendation, meaning the default namespaces do not
apply directly to attributes" part. I never managed to understand the
reasoning behind this.
Another thing that bothers me is the rough edges concerning APIs. Most
APIs handle namespaces in a really stupid way. Even XPath in XSLT
(which, IMHO is by far the best in it's anticipation of namespaces
thanks to the according axis) is incapable of dynamically producing
namespaces and one must know the default namespace to match the desired
nodes.. Also, if I'm not wrong, if there are no namespace declarations
then XSLT anticipates the empty string as the default namespace (but one
may correct me in this one). Surelly this is another rec but if
namespaces are not implemented consistently and effisiently in other
applications then there must be something wrong with them ;-)
One thing noted in other posts is URI authority but I don't think this
is important and I'm sure it's pretty tough to find a reason to use a
URI you don't own (where own means anything) as a base for yours and be
able to make something out of it.
The rest of my complaints on namespaces are in URI orbit, thus a level
lower in the architecture, i.e. what a fragment identifier denotes
(perhaps depending on MIME type), how that should be associated with an
xml:id and the like. I just don't like the document view of resources
around, which is another reason against schemaLocation vs namespaces
used for pointing to schemas or even fragments of those such as complex
type definitions (in my dreams unfortunatelly).
Cheers,
Manos
|