Add support for passing on the recipient as spamc user #1

New issue

Closed

opened 2020-03-08 13:01:46 +00:00 by glts · 2 comments

glts commented

2020-03-08 13:01:46 +00:00

(Migrated from gitlab.com)

In order to support per-user preferences, the envelope recipient needs
to be passed on to SpamAssassin with the spamc -u option.

Unfortunately, it is not obvious how to implement this properly in
SpamAssassin Milter, because of the possibility of multiple recipients.
When an incoming message is addressed to multiple recipients, one cannot
simply select one of the recipients as the user. Instead, either no
user, or a default or fallback user would have to be used in this case.
But to SpamAssassin Milter end users this will look like a bug:
sometimes my user preferences are applied and sometimes not!

Another project,
spamass-milt, uses
the above approach (its -u and -e options), which as described I
believe is flawed. So, before copying that approach in this project, we
need to think about whether that functionality is desirable and how we
want to design and implement it.

Input welcome. (I myself don’t use per-user preferences or a per-user
Bayes database, I only use site-wide preferences and Bayes data.)

In order to support per-user preferences, the envelope recipient needs to be passed on to SpamAssassin with the spamc `-u` option. Unfortunately, it is not obvious how to implement this properly in SpamAssassin Milter, because of the possibility of multiple recipients. When an incoming message is addressed to multiple recipients, one cannot simply select one of the recipients as the user. Instead, either no user, or a default or fallback user would have to be used in this case. But to SpamAssassin Milter end users this will look like a bug: sometimes my user preferences are applied and sometimes not! Another project, [spamass-milt](https://savannah.nongnu.org/projects/spamass-milt), uses the above approach (its `-u` and `-e` options), which as described I believe is flawed. So, before copying that approach in this project, we need to think about whether that functionality is desirable and how we want to design and implement it. Input welcome. (I myself don’t use per-user preferences or a per-user Bayes database, I only use site-wide preferences and Bayes data.)

glts commented

2020-03-08 14:39:50 +00:00

(Migrated from gitlab.com)

changed the description

glts commented

2022-12-24 08:45:37 +00:00

(Migrated from gitlab.com)

I will not be looking into this further, we can reopen if there is interest.