I have had the experience of being in a company where XSLT2 was *not*
allowed. Even with the XSLT1, we were told that the "advanced"
techniques should be avoided.. I was like a bit of in shock as in what
could be "advanced" in XSLT1 when the whole version is kind of  very
old now?  The reasons given to me were: This is a Microsoft based
projects and we don't want XSLT2 here.. The second reason for avoiding
the "advanced" techniques ( like keys etc ) in XSLT1 was that, the
engineers working in the MS dotNet side, wont understand it.. This was
a shame...Instead of educating the engineers, XSLT specialists were
asked to underperform in their coding ...This is true with most of the
companies where Microsoft is involved as a development platform.. It
is OK to get on to Dotnet API for XML and use dot NET file system APIs
for file outputs, instead of using for instance xsl:result-document ..

One for posterity.
*The code must be comprehensible to procedural (vb, .net c#) programmers.
* KISS for the above.
* M$ house - no java here please.

The missing item from this list?
  It's OK to shell out to VB (to do what XSLT can do) which gives
the lock in to M$ making it harder to port to XSLT 2.0
Embrace and extend still lives?

--
Dave Pawson
XSLT XSL-FO FAQ.
http://www.dpawson.co.uk