Re: XSLT in UML

bryan rasmussen wrote:

>Well XMI can be embedded in an another xml document, so I was thinking
>more along the lines as XMI as a documentation format inside the XSLT.
>  
>

or possibly one could generate a stub/skeleton xslt starting point from
XMI...the problem I have with autogenerating code from models is that
one tends to 'leave the model' early on in production...in larger
projects the benefits are much greater to maintain the model and
autogenerate code with every build iteration, though at smaller scales
(one person developer) things that take up time get dropped quickly.

In scenarios where the code that is getting autogenerated is more a
ruleset then this approach can be useful as well....though
schemas/schematron might be the right approach instead.

as an aside embedding any metadata (as u suggest XMI) to describe your
xml (in this case XSLT) is always good IMHO...though all the usual
caveats apply, namespaces may not be enough or an appropriate packaging
for such things...I am looking at the W3C compound document stuff to see
such things may work in the future on this point.

I find that there is a more mature set of XSLT users nowadays who want
to apply repeatable tests to their code (seems like a few frameworks
have emerged over the past year for test led dev in xslt), model things
properly, using a buzz word developers want to apply Agile process
within their XSLT development; we might have to borrow some modelling
tools from SCHEME/LISP to get modelling right (and lightweight) with
things like XSLT (or even perhaps other declaritive things like ant,
xaml, xul, msbuild....).

cheers, Jim Fuller