Re: [xml-dev] Using the concepts of Category Theory to create betterXML

[Elliotte Rusty Harold <elharo@ibiblio.org>]

Coincidentally I recently learned about category theory myself.

Looking over your PDF, I don't think this says anything relevant about
XML or XSLT. You're describing a very, very limited subset of XML
documents that doesn't match anything I see in the real world. What
this is describing is a very limited data model that happens to be
encoded in an XML format; but that data model is in no way limited to
or intrinsic in XML. Everything you say here could be said without any
reference to XML, nor does it seem that this approach illuminates XML.

At its base, your analysis makes one of the most common and
fundamental mistakes I see in the literature. It attempts to define
XML as just like some other well-known field (Pick your poison here:
objects, relational data, tables, categories, HTML, etc.) rather than
meeting it on its own terms and analyzing what it itself is. If I may
quote from one of my own books, with apologies to John Godfrey Saxe:

 One night five developers, all of whom wore very thick glasses and
had recently been hired by Elephants, Inc., the world’s largest
purveyor of elephants and elephant supplies, were familiarizing
themselves with the company’s order processing system when they
stumbled into a directory full of XML documents on the main server.
“What’s this?” the team leader asked excitedly. None of them had ever
heard of XML before so they decided to split up the files between them
and try to figure out just what this strange and wondrous new
technology actually was.

The first developer, who specialized in optimizing Oracle databases,
printed out a stack of FMPXMLRESULT documents generated by the
FileMaker database where all the orders were stored, and began poring
over them. “So this is XML! Why, it’s nothing novel. As anyone can see
who’s able, an XML document is nothing but a table!”

“What do you mean, a table?” replied the second programmer, well
versed in object oriented theory and occupied with a collection of XML
documents that encoded UML diagrams for the system. “Even a Visual
Basic programmer could see that XML documents aren’t tables.
Duplicates aren’t allowed in a table relation, unless this is truly
some strange mutation. Classes and objects is what these document are.
Indeed, it should be obvious on the very first pass. An XML document
is an object and a DTD is a class.”

“Objects? A strange kind of object, indeed!” said the third developer,
a web designer of some renown, who had loaded the XHTML user
documentation for the order processing system into Mozilla. “I don’t
see any types at all. If you think this is an object, then it’s your
software I refuse to install. But with all those stylesheets there, it
should be clear to anyone not sedated, that XML is just HTML updated!”

“HTML? You must be joking” said the fourth, a computer science
professor on sabbatical from MIT, who was engrossed in an XSLT
stylesheet that validated all the other documents against a Schematron
schema. “Look at the clean nesting of hierarchical structures, each
tag matching its partner as it should. I’ve never seen HTML that looks
this good. What we have here is S-expressions, which is certainly
nothing new. Babbage invented this back in 1882!”

“An S expression?” queried the technical writer, who was occupied with
documentation for the project written in DocBook. “Maybe that means
something to those in your learned profession. But to me, this looks
just like a FrameMaker MIF file. However, locating the GUI is taking
me awhile.”

And so they argued into the night, none of them willing to give an
inch, all of them presenting still more examples to prove their
points, none of them bothering to look at the others’ examples.
Indeed, they’re probably still arguing today. You can even hear their
shouts from time to time on xml-dev. Their mistake, of course, was in
trying to force XML into the patterns of technologies they were
already familiar with rather than taking it on its own terms. XML can
store data, but it is not a database. XML can serialize objects, but
an XML document is not an object. Web pages can be written in XML, but
XML is not HTML. Functional (and other) programming languages can be
written in XML, but XML is not a programming language. Books are
written in XML, but that doesn’t make XML desktop publishing software.

XML is something truly new that has not been seen before in the world
of computing. There have been precursors to it, and there are always
fanatics who insist on seeing XML through database (or object, or
functional, or S-expression) colored glasses. But XML is none of these
things. It is something genuinely unique and new in the world of
computing; and it can only be understood when you’re willing to accept
it on its own terms, rather than forcing it into yesterday’s pigeon
holes.

-- 
Elliotte Rusty Harold
elharo@ibiblio.org