Avoding a repeat of W3C XSD - was Re: [xml-dev] Is Web 2.0 the new XML?

[Michael Champion <michaelc.champion@gmail.com>]

On 8/16/05, Michael Kay <mike@saxonica.com> wrote:
> > No. We need XML Schema so that other WGs know what not to do.
> 
> OK, so what exactly did they do wrong, and how can other WGs avoid doing the
> same?

That seems like a useful question to ask, especially since W3C seems
to be trying hard to learn from what XSD and some WGs did wrong.

My strawman list would start with:
- Trying to do too much at once with a single spec. There's some
famous quote along the lines of "all successful complex applications
were once successful simple applications" that we should all probably
ponder daily. I wish I could find it ... it probably came from Sean
:-)
- Not having a rigorous testing process in place BEFORE Recommendation
(the W3C "Candidate Recommendation" phase and other process changes
address this)
- Not appreciating that the formalism has to come first (as RELAX NG
was built on the hedge automata formalism) to be of real benefit. 
(XQuery at least tried do this in parallel, with some success)

There's also the uber-problem that all WGs have, which is that the
consensus process makes it much easier to add stuff than remove stuff.
 I don't know how to address that, but as I recall the Schema WG
*explicitly* made it hard to remove features once they had been agreed
to.