Achim Blumensath writes:

 > use four instances, one for each of these cases? That's the approach my
 > one head-package used (which never made it to the experimental
 > directory, but it should be somewhere in the mailing list archives).

adding Peter's contribution was not meant as a value statement one way or the
other. so if you dig up your attempt i'll put it side-long (or anything else
for that matter as long as it is documented and helps in thinking further
about these matters)

what the disucssion so far tells me (though that isn't really surprising :-)
is that what is really missing is thinking further about the interfaces
itself, e.g.

  - what kind of template-type (s) is wanted, what exactly should it do?
  - what are the explicit/implicit inputs?

frank