[topicmapmail] A somewhat new topic maps format

[Alexander Johannesen]

Lars Marius Garshol wrote:
> [...] but it would be good if you posted a converter
> from CSXTM to XTM, so that people who use it are not trapped in a
> non-standard format.

Of course, and it is coming up.

...

> <URL: http://www.ontopia.net/download/tmstrict.dtd >
>
> for Ontopia's version of the HyTM DTD. (It's an architectural form,
> which is why different vendors have different versions of it.)
>
> The benefit of using this is that it's near-standard, pretty well
> designed, more compact than CSXTM in several ways, and already in
> use.

I'll look into it. My first question is if it is optimised for
XSLT? CSXTM was, and no other technology was thought of when doing
it. I honestly haven't looked into HyTM as I thought it was a
format somewhat abandoned (even if I know OKS supports it; any
other?)

Thanks for the comments on the format:

> Your examples have -id-s which begin with numbers; that's not allowed in 
> XML.

Another example of examples that one thought would be simple and
intuitive, really clouds and is plain wrong. :) Will fix.

> There's no point in trying to preserve the -id-of a <scope> element.
> The data model does not preserve these, and so they are not really
> part of the topic map.

My thinking here is that since XTM allows them, some might use it,
and as you've already pointed out, the missing CSXTM to XTM is
coming up. They are meant to be at least compatible on the basics.

(On a side note; should this rather be done for the upcoming XTM 1.1?)

> A scope is a *set* of topics, but it doesn't seem that your -scope-
> attribute can handle this. If so, that's a problem you really ought to
> fix.

Yes, at the moment there is a one-to-one scoping. I'm thinking
that this could be another case of <also scope="..." />.

> Variants are required to have a scope, but your syntax does not seem to 
> enforce that, though it should. If this is what -refid-is supposed
> to do I'd replace it by a -scope-attribute (and sub-element), since
> this is what the data model maps it to, and it would be easier for
> people to understand.

Agreed; refid becomes scope, indeed.

> There's no such thing as a "published subject identity", I'm afraid,
> and from your description it's not clear how the <psi/> element is
> meant to be used. It would be good if you could clear that up.

It is a typo. "published subject *Indicators*". Will fix. Other
subject indicators will come later, but this was the one I needed
right away.

> As regards associations you've left out -scope-there.

Yup.

> Members, on the other hand, only exist in the XTM syntax. Elsewhere
> they are called association roles, and I would strongly recommend
> using that term instead.

Why? As CSXTM was meant to be a simplified version of XTM, wouldn't
"member" fit the bill?

> Finally, association roles (or members, if you like) do not have
> scope.  It's the associations that have scope.

You're right, but while doing this it *felt* like there should be one. :) 
It will be moved up to association where it was meant to be.

> Oh, and one more thing: it would be good to have a DTD for this. Prose
> just cannot compete with a DTD for conciseness and precision, and you
> can't validate an XML document against prose.

The prose is because of the incredibly swiftness this was done in. :)
If anyone really are interested in this format - i.e. it survives
the test of actual implementations - I might do so. :)


Regards,

Alexander
-- 
___________________ ______________________ _____________________________
                   |                      |
http://shelter.nu/ | alex at shelter . nu | http://shelter.nu/xsiteable/
___________________|______________________|_____________________________