Re: [xml-dev] A question of necessity

[Alexander Johannesen <alexander.johannesen@gmail.com>]

Arjun Ray <arjun.ray@verizon.net> wrote:
> You may have misread mine.

Not impossible.

> In what ways would treating colonified names as atomic names have
> made things difficult for you?

What's an atomic name in XML? Do you mean, treat it as any other
element with no special treatment? I use a lot of XSLT where the
namespace schema declaration is what changes depending on context
(different versions of different ontology languages, essentially), and
then we merge a lot of different ontologies in a mixed content model.

For example;
   <nut:template xmlns:nut="http://schema.shelter.nu/nut/v1";> ...
</nut:template>
vs.
   <nut:template xmlns:nut="http://schema.shelter.nu/nut/v2";> ...
</nut:template>

The difference in versions (and there's average about 20 different
ontologies with a few good versions each) renders the result very
different, as some ontologies aren't compatible in some versions of
others, and so on. There's a small logical model that parses out the
various versions and compatibility, and the XSLT layer renders the
rest.

But again, necessary? In the way the system is designed, yes. But it
can be designed differently, as always. So I'm still wondering what
necessary mean. :)

> Sorry, I have no idea what you're talking about.

Fair enough.

> Please define, or at least clarify, mixed content models and sanity
> thereof.

<nut:upper name="object">
   <smic:middle name="passive">
      <core:object>
         ... properties ...
      </core:object>
   </smic:middle>
   <smic:middle name="active" /> <!-- parsed as original -->
</nut:upper>

I "speak" the NUT and SMIC ontologies fluently, so when I create the
core ontology (core namespace), I wrap them up like this. Then I can
try out various versions to find out how it affects the various
properties of the resulting ontology. For each domain there's a
generic stylesheet for parsing, and one for version rules.

The schema versions to each namespace will render the resulting merged
ontology subtly different. Easy to do in XML and XSLT, but without
namespaces I suspect more backend coding. I love the functional nature
of XSLT, which is why I chose this solution, and I'd rather not resort
to more common imperative languages unless I have to, but hey.

Again, can I do it with plain names? Sure, but very, very differently.

> That is, the necessity of "universal names" was simply laid down by
> fiat.  I still don't think that was good enough.

Hang on. Are you opposing namespaces, or the ruleset implementation of
namespaces in XML? There's rules in it I certainly don't agree with,
but the overall idea certainly fits me well.

Cheers,

Alex
-- 
 Information Alchemist / UX badass / GUI wrangler for hire
 http://thinkplot.org     |  http://www.linkedin.com/in/shelterit