Reply-To set to Mailman Developers. Please check that your replies go there. Or you can comment on MRs directly if you have a GitLab login.
I have three merge requests pending:
[1] https://gitlab.com/mailman/mailman/-/merge_requests/1159 [2] https://gitlab.com/mailman/mailman/-/merge_requests/1160 [3] https://gitlab.com/mailman/mailman/-/merge_requests/1161
[1] fixes the logging initialization bug that prevented the smtp
log
from rotating properly, and maybe some others. I don't think it needs
user review because it's a pure bugfix, but if you're curious and take
a look, and you think it needs better documentation let me know.
[2] adds a configuration variable lmtp_max_size
. This is
user-visible, so review of the documentation in docs/mta.rst and
config/mailman.cfg for clarity and readability would be appreciated.
[3] adds postfix-pgsql support for use of the PostgreSQL database to route Mailman traffic. It's really most useful in an edge case (migration of 10,000+ list Mailman 2 instances), but it's theoretically attractive on the principle of "single source of truth", and possibly for implementing a "microservices" style cloud instance (it eliminates the need for Mailman and Postfix to share a file system). Besides opinions on the value of the general idea I have two specific questions:
- Postfix docs recommend not looking up domains in the database to reduce load on the database server. But on a dedicated Mailman system, you're going to be looking up pretty much everything twice anyway (once to check for local recipient and once to fetch the transport), and my experience so far is that this isn't a big load on Postgres, and it's not a drag on list traffic distribution. Should I implement domain lookup in Postgres? Should it be optional if the lists are being looked up in Postgres?
- The current patch completely disables generation of the
postfix_lmtp alias file, even for the
mailman aliases
command. Should I implement an option to force generation of the files for documentation (and maybe a backup in a very rare and contorted situation where Mailman can see Postgres but Postfix can't)?
Thanks, Steve