path-searching again

[David Carlisle]

Pierre> I'll try it.  I'll put a tfm in that position, and my
Pierre> expectation is that the worst I will get is a single path
Pierre> through pk. 

But that is exactly what was at the crux of the months of discussion
we had over fonts/type vs fonts/source. Although efficiency is an
issue on some systems, Karl pointed out 1000 times that given a decent
caching algorithm there was no real efficiency concern here. My
objection was always that fonts/source did not allow you to *specify*
where TeX or anything else should look for tfm files.
texmf/fonts//tfm still specifies a search over all of texmf/fonts.

Irrespective of efficiency, it seems a natural wish to be able to
specify that the tfm input path should be over directories containing
tfm files, texmf/fonts/tfm// seems to be the only way of doing this
with full recursive directory searching. Otherwise you would have to
need one-level wild cards and make use of the particular arrangement
of the TDS tree, as in

texmf/?/?/tfm

(or however many levels deep it should be:-) Such a method would
again cut out the pk and source directories, but at the price of a
much more rigid structure.

David

Why am I being sucked into another fonts/source discussion!