It's only used by the string-contains[-ci] procedure, and that
   procedure may as well be implemented using Boyer-Moore.

You can't use Boyer-Moore with large character types -- it requires you
to build a table with one entry for every possible character. Hence
not really useable for anything past Latin-1.

In fact, I have an implementation of B-M. I just don't *export* it into
the library's API as such, because it isn't portable across character types,
which is one of the design criteria of the lib.
    -Olin