Re: breaking up?

[Rick Jelliffe <ricko@allette.com.au>]

 From: <Mike.Champion@SoftwareAG-USA.com>
 > > From: Michael Brennan [  <mailto:Michael_Brennan@allegis.com>

> > Too bad we can't go back and start over with the benefit of hindsight.

> "Just say no" to defaulted attributes,
> ambiguous namespaces, validating with schema languges that are harder to
use
> than to write procedural business logic, and so on.

> If the folks "leading the web to its full potential" via the PSVI-oriented
> specs turn around in a couple of years and discover that no one is
> following, why is that a problem for the rest of us?  Conversely, if they
> *do* sort it all out and make it work in the real world someday, what have
> we lost by letting them do the bleeding on the bleeding edge?

Hear, hear.

But I think there is another wrinkle to this.  We are talking about how
good a technology or architecture is (e.g. PSVI, W3C XML Schema),
but many of the big players will be thinking in terms of products and
strategies.  The big player will want its users, ultimately, to purchase its
products and will be proclaiming "We can shield you from the
complexity with our nice GUI tools and seamless, total-solution,
register-once/pay-forever services" or whatever.   The complexity
of doing it low-tech creates the market for "shielding" solutions.

So I don't think we should expect much leadership from companies
who sell end-user schema tools.  Sun's recent multi-schema
validator is one of the few bright spots, I think. Even though
it is taking the SGML kitchen-sink approach, it can provide a really
nice basis for people writing their own grammar-based schema
languages or who want to use RELAX NG.

If people don't like XML Schemas and don't want to buy into a vendor's
line, there are currently viable alternative approaches:

  1) Examplotron is so simple and straightforward for small messages
I find it difficult to justify anything else.
  2) Schematron has an unbeaten power/performance tradeoff (IMHO)
for open schemas, complexe co-occurrence constraints, and non-regular
constraints.
  3) XLinkit is a high-level language for servers which is suitable
for disciplined software engineering QA, sort of a super-Schematron.
  4) RELAX NG is going full steam ahead, and should be very efficient,
with several implementations.
  5) Simon St L's Regular Fragmentations can allow validation of
complex embedded notations quite easily.

On top of those, I think there are two other avenues which would be
worth investigating:
  1)  A minimal subset of W3C XML Schema's structures.
  2)  An alternative syntax for RELAX NG or W3C XML Schema
that ramps up DTD's content model syntax.  (There was one floating around,
it is unfortunate it has sunk.)  For example, to use
  {1 to 3} for occurrence
  [ a, b, c] for all groups
  x:*  for any in a wildcard in x namespace
  *:*  for any in any namespace
  ~:*  for any in no namespace or whatever.

Having an alternative syntax would do quite a lot to endear
W3C XML Schemas (or RELAX NG) to people IMHO.
The diffulty of reading/writing XMl Schemas in eleemnt
syntax shows how bogus the idea that all structure should
be (only) in element syntax is. (Just as bogus as the idea
that no structure should be in element syntax, no flames please.)

Cheers
Rick