[
Lists Home |
Date Index |
Thread Index
]
james anderson wrote:
> the programming problem is an artifact of the misconception that names in a
> namespaced world need be modeled at the application level as tuples. this is
> no more true than would be the claim that names in a non-namespaced world need
> to be modeled at the application level as unicode scalar value tuples.
The programming problem is an artefact of the XML Namespaces spec.
I need three things to define a name with namespaces, a namespace
name, a qname (which is arguably an embedded tuple), an xmlns
declaration. There isn't any other specified way to do this, what
happesn at the applciation level is a second order effect.
>>In other words namespaced elements are tuples after a kind.
>
>
> if one insists on modeling them that way, one is making ones own problems.
The 'problem' lies squarely in the Namespaces spec.
> cl-xml includes a first cut xml-path processor. it adds to the standard
> parsing protocol that xml path expressions are interned in the parser's
> dynamic context. from that point on, the only aspects of the processor - and
> therefor of application code, which are concerned with lexical properties
> (prefixes, namespace names, local parts) are those which the xpath standard
> stipulates must be present - eg to match on local part or on namespace name only.
>
> everything else can be expressed in terms of abstract names. no tuples. no
> lexical properties.
To use XML Path I need an xpath and an xml document first and
foremost. What happens when evaluating an xpath with a Qname depends
on which namespaces are in scope via the namespace nodes for that
element and the expanded name for that element, but the point is
that the xpath string needs to declare upfront whether it's testing
for a Qname or not and the processor needs to know whether the in
context element has an expanded name. Heck even the XPath spec
defines the expanded name as a tuple. What you're describing sounds
like a processor for something other than XPath, possibly an
XPath+macros processor - if I'm missing something perhaps an example
will enlighten me.
>>In turns out that you can't without incurring a development cost.
>
>
> true, but only in that one would need to fix what appears to be an
> inappropriate document model.
Strawman objection. This is a not about finding the right document
model for XML Namespaces, it's about how XML Namespaces models XML
names. If we're to going have to tease apart a conflation of ad-hoc
document models with a specification model for XML names, we should
call it a day, cos we'll be here all weekend talking past each other.
>>There is no canonical (self-describing) form for for a namespaced
>>element,
>
>
> it must be that i do understand what this is intending to claim, as, as i do
> understand it, i have thousands of lines of code lying around here which
> disproves it.
I looked in the XML and XML Namespaces spec and couldn't find such a
form. Then I looked in the Infoset, just in case. I couldn't find it
there either, but I did find plenty of tuples. I found a way to
infer what the name is in terms of it being distinct from other
names (essentially how expnaded names work in XPath), but I can't
easily write it down or inscribe it, say on a piece of paper, though
I'm can use James Clark's {} idiom. Really, I'm interested in the
Namespaces specs model of an XML element, not cl-xml's object model
of the Namespaces's specs model of an XML element. The point is
this: I don't want an encompassing OM (even if it is in clisp) to
protect me from the details of imho a cost-ineffective standard, I
want the standard exposed, in the hope that people will stop
suspending their technical judgement when it comes to using
namespaces in the first place.
Btw, I went to http://www.cl-xml.org/ - how do I download cl-xml?
>> - technically the XML grammar won't allow it.
>
> please amplify. it is not clear how the grammar places this constraint on the model.
URI references can contain characters not allowed in names, so
cannot be used directly as namespace prefixes. Therefore, the
namespace prefix serves as a proxy for a URI reference. Prefixes are
a workaround - they're not even syntax sugar.
Bill de hÓra
|
