Thanks very much for digging into this, Michael.  A few points:

First, we at Microsoft acknowledge that we erred in submitting the tests wi=
th non-conforming regex expressions to the W3C.  A decision was made some t=
ime ago to align the regex behavior in the MS XSD implementations with that=
 in the underlying Windows and .NET libraries.  Naturally our internal test=
s are aligned with our specs, but we should have culled these out before su=
bmitting them to W3C, and apologize for this oversight.  W3C should probabl=
y simply remove all those  tests that exercise regex functionality for whic=
h you filed bugs.

Second, we now have our critical mass of XSD experts back from the holidays=
, and we will dig into the other cases you flagged. We commit to publishing=
 our report on conformance with the W3C suite at the same level of detail t=
hat you have. Those reports which we accept as bugs will be filed against o=
ur implementations, and those about which we dispute your interpretation of=
 the spec will be discussed publicly on xmlschema-dev and/or the bugzilla c=
omment threads.  We hope that the result will be both more conformant imple=
mentations from all of us and (probably) a list of ambiguities in the XSD s=
pec that the WG may wish to consider in drafting 1.1.

Third, it may be worth noting that we ran the Sun and NIST tests when they =
were first submitted to W3C and filed two bugs against our implementations.=
  So overall, our implementations are 100% conformant with the tests we sub=
mitted and more than 99% conformant with the others.  That, at least in my =
humble opinion, indicates that XSD interoperability is pretty good at this =
point.  As you mentioned, it is mainly corner cases where various implement=
ations give different results.

Finally,  about our processes for fixing the bugs in our implementations an=
d tests that are exposed by this exercise: In general, we promise our custo=
mers to NOT make "breaking changes" customers in service packs, so don't ex=
pect any bugs that we acknowledge to be fixed before the next release of th=
e MSXML or System.Xml libraries.  I can't commit to all conformance bugs be=
ing fixed in the next releases, but we will give these very careful conside=
ration.

Michael Champion
Program Manager, XML Standards
Microsoft Data Programmability - XML team