Re: String hashing code

Robert Koberg wrote:
> On Sat, 2007-12-15 at 11:31 +1100, Deborah Pickett wrote:
> Just make the path from some root dir the filename (replacing the
> slashes with whatever) and then append generate-id() or position()
> (possibly combined with depth and prev-sibl count). You could create the
> file with a unique extension (*.mathml ?) for easy fileset selection in
> ant. Also your main output can create a list of all the files.

Yeah, that's sort of what I've done.  It's not ideal, because the path
names are very long and likely to break the 255-character limit on path
segment names in the Windows filesystem.  The final n (I picked 30)
characters of the path are probably unique, but I'd feel more confident
if every character in the source filename contributed to the uniqueness.
   Or some other mechanism which doesn't need the source filename to
generate unique-enough-ness.  (Which is where I came in with my original
message.)