> if the tree contains only files of interest, then VMS lib$find_file
 > will certainly be fast and efficient;
 > if the tree contains irrelevant files, then lib$find_file may well
 > be too slow.
 > Which is the situation for TDS?

well, if you take the TeX Live CD, there are 213 LaTeX packages with
at least 1 file, maybe several. Say 500 files; if
you do \usepackage{foo}, at a minimum you have to scan the tree
containing those 500 files to locate foo.sty. Is that good or bad?

Then again my CD has over 5000 tfm files. If I do
you are going to have to look at all of them, arent you? hard to see
how to cut that down rationally.

Whether the tree has irrelevant files depends, of course, on how well
you specify the subtree to look at. Probably someone like the AMS
would set up a `vital' tree, and a `last resort' tree so that
something like cmr10.tfm was located fast.