Richard and Joe--

> Interesting project!

I think so :-).

Thanks for the ideas.  I guess I could use xslt (which I am using
elsewhere in the project) to handle the inclusions, but it seems like
overkill at this point, when all I want to do is pull in some
boilerplate.  But I'll think about it--there is some pre-processing,
although at this point it's done by Python, since there is some fairly
heavy looping involved (producing all possible combinations of a set of
sets of features).  In fact the only reason for using xmllint was to
ensure that the file Python was going to read was valid (thereby making
it easier to debug).

The actual generation of inflected words is done by an entirely
different program, the Xerox Finite State toolkit, which converts forms
like compl+1sgS-3sgO-pale:wia-perfv.sg into o:+n-k-pale:wih-0 by
replacing glosses with Nahuatl affixes, and thence into o:nihpale:wih
by applying phonological rules.  I don't think that's a job for xslt
:-).

Anyway, thanks to both of you for all the ideas, and the helpful
discussion!