RE: ZDNet Schema article, and hiding complexity within user-friendlyproducts

[Michael Champion <mike.champion@softwareag-usa.com>]

> -----Original Message-----
> From: Murali Mani [mailto:mani@CS.UCLA.EDU]
> Sent: Monday, April 23, 2001 9:27 PM

>
> SQL has a nice math based on relational algebra/calculus.
> The truth is XML Schema does not have this.
> Actually probably because of my strong bias, but I believe XML Schema
> defies math -- for an example, XML Schema defies XDuce.
> But this is true -- the math for RELAX/TREX was first studied in the
> late 1950's, and it has *very* strong math.

Interestingly, the Schema Working Group is/was? trying to develop a
formalism upon which to base the syntax.
See http://www.w3.org/TR/2001/WD-xmlschema-formal-20010320/
There are a number of "infidelities" between the spec and the formalism,
however. For some reason, the WG appear to have decided that these are
bugs in the FORMALISM rather than problems lurking in the spec. (The
unsanitized formalism draft at
http://www.wavelet.org/cm/cs/who/wadler/papers/msl/msl.pdf
calls them "problems with XML Schema", the public draft on the W3C site
calls them features that are "not currently modelled.")

As depressing as it is to be reminded of all this <grin>, I think you've
hit upon an important answer to my original question: User-friendly
interfaces can hide TEDIOUS FORMAL DETAILS of a spec that no sensible user
would ever want to touch (e.g. a fax machine hides the CCITT image
compression specs from the person pressing the buttons, and SQL hides the
formalisms of the relational algebra). But it's not at all clear that they
can hide conceptual complexity (and perhaps ambiguity) such as the XML
Schema restriction rules. To quote from the (un-sanitized) schema formalism
paper referenced above, "An attempt at defining formal rules corresponding
to the definition of restriction required more than 100 lines of formal
rules (as compared to the one-line definition given in the current
document). Many rules had five or six premises."   Something tells me that
we're not going to see a clean, user-friendly interface that encapsulates
these 100 rules from Microsoft or anyone else.