Lists Home |
Date Index |
"Thomas B. Passin" <email@example.com> wrote:
| Namespaces alter the generic pattern but are still pretty generic.
I find this statement quite inscrutable. In what sense are namespaces (as
defined in the Rec) "generic"?
I said "generic" because namespaces apply to all the specializations (at
least, those that make use of them), but the details of the namespaces and
what they denote can be specialized. This is much like providing for
element names in xml of sgml - the generic spec sets forth hwo to construct
names but leaves it to the specialized uses to refine and constrain.
Even though I labeled namespaces as a generic addition to the generic nature
of the original xml rec, I certainly agree that they constrain that rec. I
am saying that they do it at a fairly non-specific level.
What is generic about a device that does not allow values to be shared by
names from different taxonomies? (As in, why had it to be xlink:href
*versus* html:src?) What is generic about an approach calling for extra
elements without any determinate notion of opacity, that also need to be
kept away from pretty printers lest the whitespace gum up the "careful"
coding (tags jammed together to box the shared value)?
| It is like tightening the tolerances on screws.
No, it's like saying that screws must work for all materials.
I think you are articulating one of my points, that the way more specialised
or contraining specs work with the more generic ones is very important but
also it is non-trivial to arrive at a good solution.
Continuing the analogy I introduced in my earlier post, it might have made
sense to have namespaces be a augmenting spec or profile that could be
picked up by those who need it but not everyone. With SAX we can use it
that way (since we do not have to use namespaces), and almost with xslt (we
have to use namespaces in the stylesheet but not in the document being
transformed). This seems pretty good to me. The only thing we really
cannot do is to use a colon in a name because most parsers will assume that
namespaces are supposed to be in effect.
So I would argue that, from the point of view of augmenting the basic
generic spec (xml 1.0), namespaces did pretty well. This is completely
different from accepting the engineering design, BTW. Everyone can decide
for themselves if XML Schemas, XPATH 2.0, and xslt 2.0 have that same