Well, my hobbey horse is that it is mistake to think that any but the most trivial of specifications/standards can be implemented outside of a community which includes some judicial function to resolve obscurities and debates. In the case of XML Schemas, that presumably will be XML Schemas 1.1 and beyond (and it perhaps may not be utterly inconceivable that some people may tend to think, all things considered, that it has more obscurities and debates to be resolved than many other specifications may appear to have, for better or worse.)
But even if I think the expectation of perfection is naive, I think Mike is correct that there needs to be a feedback process. I don't think perputual revision is possible with W3C specs and their ilk, but I don't see why there cannot be a "staged revision" system, where a new revision is released every year or three with an effectivity notice that it will take effect at a certain time.
Anyway, surely one big thing that stands in the way of perfection is monolithic specs, where we cannot upgrade the bits that prove troublesome or deficient in practise. A proper modular architecture, where names/namespaces, grammars, datatypes, non-regular constraints, etc. were independently upgradable.
Actually, not all the implemented the working group tracked were created by people who had the privileges of being a Working Group (or even an IG member), I believe. There was discussion in the WG about whether XML Schemas should even be released without a test suite, conformance language, formal model, etc. However, since no other spec had started its public life that way, and since it would add at least a year before release, and since it had not been something in the requirements documents that people should expect, and since a conformance suite may be seen as something which (in its early days) provides an alternative source of what a specification means (contrary to the principle that you should only specify something once in one place, if possible, because of maintenance and consistency reasons) it was decided to leave those jobs till after the rec was out.
From Mike.Champion@SoftwareAG-USA.com <Mike.Champion@SoftwareAG-USA.com>
SHOULDN'T there be some stage between "We've demonstrated that people on the WG can implement the spec" and "We've seen that independent developers can implement the spec solely from the information it contains, and those independently developed versions are interoperable".