[Mailman-Developers] Re: file permissions

[Barry A. Warsaw]

Normal Friday catchup...

>>>>> "MM" == Michael McLay <mclay@nist.gov> writes:

    MM> Either I screwed up when I followed the installation
    MM> instructions or else the file ownership set by "make install"
    MM> incorrectly set the group of the executables in
    MM> $prefix/cgi-bin to that of the super user.  In order for the
    MM> scripts to work the group had to be manually set to "mailman".

One thing the next version will have is more checking of the
permissions on $prefix.  If the permissions or group ownership aren't
correct, configure stops.  This should prevent those problems from
cascading to all the other installed files.

    MM> This change should also reduce the script execution time
    MM> slightly because the longer script will be compiled to a .pyc
    MM> file.

That may or many not really be an issue.  I suspect starting up the
Python interpreter for every script is the killer, rather than parsing
the script files.  subscribe for example is less than 200 lines of
Python.  Admin is certainly the longest at 856 lines.

One thing that's high on my list is to really fix the packagization of 
Mailman.  I think given that, many things get simpler.  You still need 
a little path hacking, but that can be done in a more portable way
than even the paths.py files.  And then most of the functionality in
the scripts can be moved into package libraries.  This will improve
the modularity A LOT.  I just finished doing this with a project here
at CNRI and it makes a big difference.

    MM> John, what do you think of removing the mm_ prefix from the
    MM> module names?

This can easily happen, with no chance of name collisions, even for
the mixin classes.  I also want to rename some of the .py files to
match the contained class names (better style).  So there are several
things of this nature I want to work on, but my meta question is this: 

From a technical standpoint, how close to a 1.0 release are we?  What
critical features are still missing (integration with Pipermail is
key), and what showstopper bugs are still lurking?  Should I continue
to do code clean up now, or wait until after 1.0?

-Barry