Adam (01)
> I appreciate the detailed and concrete response. While at some level I
> agree with you, I feel that we may be approaching a religious
> debate. (02)
Hopefully not. Too often religious debates lead to all out war :-( (03)
>Certainly, with a lisp reader one could have the same sort of
> hierarchical data structure and lisp readers would tokenize the keywords
to
> save space, so that's not unique to XML. But at that point, we're talking
> about internal data structures, and not the interchange language. XML has
> virtues because of it's current popular acceptance, but there are still
> many issues that would need to be resolved. (04)
I'm not claiming it as a panacea, just as an approach that allows for
declartive processing, which is something I happen to like. (05)
> Your element type approach does yield a more compact structure, but it
> appears that you have lost the full expressivity of KIF in doing so. For
> example, KIF allows a variable, terms, or function as an argument and your
> proposal for the form
>
> <instance id="variable-X">
> <type>Horse</type>
> </instance>
>
> seems to assume that neither argument is a function application. (06)
I have not begun to attempt to fully model KIF. I simply tried to show that
the example of XML given was unnecessarily verbose, as most examples I see
from people who want to make the point that XML is verbose. I was trying to
kill a chestnut that should never have been thrown on the fire. (07)
> I'm playing devil's advocate a bit here, but what's the compelling
> benefit of this syntax change, even if you create an XML encoding
> sufficient to handle the full expressivity of KIF? (08)
I'm always the devil's advocate, but the one pushing XML as the diabolical
art. :-) (09)
> The interchange
> language becomes more verbose, and any tools that handle generic XML would
> need to be specialized to handle not only the syntax of XML, but the
> particular features of the encoding, and the semantics of KIF. (010)
No, that's my point. You don't need a specialized tool. You simply devise a
set of declarative statements that allow a generalized XSLT tool to convert
your XML representation of KIF to the required language. Yes, you can do
that with any LISP compiler as well, but do you get that free with your
operating system, in the way you get the XSLT tranlator? (011)
> So you've
> just added another layer of software structure. (012)
You have to have another layer, whether its XML, LISP, Perl or Java. You
need to be able to transform one strucuture into another. Its better to do
this with a tool designed with such transformations in mind rather than with
a general purpose tool. (013)
But I don't expect to convert you. Sometimes you have to let the heathens
keep their own religions and simply use your own works as an examples to
others of how they can improve their lives. (014)
Martin (The XML Devil's Advocate) Bryan (015)
--
To post messages mailto:ontolog@xxxxxxxxxxxxxxx
An archive of the [ontolog] forum can be found
at http://ontolog.cim3.org/forums/ontolog (016)
|