Stephen J. Turnbull
stephen at xemacs.org
Mon May 4 02:15:34 CEST 2015
Mark Sapiro writes:
> On 05/03/2015 06:17 AM, Laura Creighton wrote:
> > On all the machines I can ssh to, I have an
> > /etc/mailman/mm_cfg.py as well as a /usr/lib (or /var/lib)
> > /mailman/Mailman/mm_cfg.py, and the second is a symbolic link to
> > the first. Why do we have an /etc/mailmand/mm_cfg.py at all?
> In a word (or acronym), FHS.
> Actually, AFAIK, the Red Hat (hence CentOS) Mailman package is the only
> one that does it right.
I don't understand what you mean by "right".
John Dennis's distinction that mm_cfg.py is "executable" and others
aren't is bogus AFAICS; there are lots of apparently data-only files
that are actually interpreted code (PDF being one of the more powerful
What matters here is exceptional treatment of this one file by the
PMS. In the case of Red Hat, they cannot simply unpack the archive
into place (in practice, noboby does that any more, but that's a
different issue); they have to either avoid copying over mm_cfg.py or
preserve it and restore it. In Debian, you can just unpack the
archive, but then you need to remove the distributed mm_cfg.py and
restore the symlink.
AFAICS these procedures are basically equivalent, except that both the
PMS and the sysadmin have to know something special about Mailman in
the Red Hat scheme. Note that in the Debian scheme you can be an old
Mailman hand and edit /var/lib/mailman/Mailman/mm_cfg.py, or Debian
admins who don't know much about the history of Mailman can just go to
the FHS place in /etc/mailman and find mm_cfg.py there. That's an
advantage to the Debian scheme AFAICS. Another (minor) advantage is
that in the Debian scheme the PMS doesn't have to do anything special
to protect the "real" mm_cfg.py, it just never automatically edits
anything under /etc.
More information about the Mailman-Users