This SRFI addresses a real problem, and I think provides a good
solution. However, I have a concern about terminology.

It has been noted before in various places that we should not assume
that libraries are in files. I like to imagine (and would implement it
if I had the time) a PL/Scheme that lived inside Postgres, where there
was a table called Libraries that contained all of the libraries. I
therefore think that this should be slightly rewritten to say that an
implementation can arrange to have a mask, with the properties given
here, associated with each library, and recommending that a file-based
implementation should look for the mask in the file mask.slm in the
library directory.

Masks are useful for implementers, as shown in the examples in the
SRFI,  but I am not sure I understand if there are any implications for
the developers of individual libraries. I will need to think about that.

-- vincent