Re: MAIL_USE_FLOCK and Debian.

[Michael Sperber [Mr. Preprocessor]]

>>>>> "Richard" == Richard Stallman <address@hidden> writes:

Richard>     Given the discussion so far, it seems like we could:

Richard>       - provide a run-time choice between the flock/lockf and maillock
Richard>        approaches on systems that support both, along with a
Richard>        configure-time choice between the liblockfile and libmail
Richard>        implementations of maillock when relevant.

Richard>       - provide a run-time choice between flock/lockf, maillock via
Richard>        libmail, and the liblockfile algorithm.  With this approach, we
Richard>        wouldn't use liblockfile itself (because its maillock would
Richard>        conflict with libmail's), just the code.  Though direct use of 
the
Richard>        code definitely has potential maintenance issues, at least as 
far
Richard>        as Debian policy is concerned, I believe that just implementing
Richard>        the algorithm is acceptable.  Of course, if we did use the code
Richard>        directly, we'd need copyright assignments.

Richard>       - build-time choice only -- similar to what we have now, but 
perhaps
Richard>        with an added --with-mail-locking=FOO which would require a
Richard>        specific approach.

Richard> The second is a real pain.  Do you really want to go to that much
Richard> trouble?  If not, our choices are the first and the last.

Richard> Would you like to do the first one?

XEmacs movemail does the first and third.  (It allows you to specify
a compile-time default and then override it at run time.)

I've offered cooperation on producing a shared movemail previously.

-- 
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla