Enhancement: Make parsing regexes configurable

[modoboa]

Enhancement: Make parsing regexes configurable

[tonioo]

Indeed. How would you configure it? Through settings.py or through online parameters?
Do you have an example?

[mundschenk-at]

Good question. I think several preconfigured variants selected by a single choice in the online parameters would be most user-friendly. Optionally, customized regexes could be set via settings.py. I'm working on proof-of-concept PR (probably without the settings.py part).

[mundschenk-at]

@tonioo I've added preliminary Rmilter support in PR #19. To have complete support (including virus scanning), multiple Rmilter log lines would need to be parsed. This would probably only make sense with a complete revamp of the _parse_line method.

What I'm currently unsure is the intended specification. Are spam/virus mails counted separately from sent/rejected mails? If so, we don't need to do anything. If spam mails should not count as rejected mails, we would need to have more complicated logic to prevent processing of the same message ID from another prog.

[tonioo]

@mundschenk-at I think that spam/virus should not be counted separately because we don't know the configuration (eg. if traffic is going to be rejected or accepted). Does it answer your question?

[mundschenk-at]

@tonioo That would seem to be a contradiction. The lack of information about what to do with spam and virus mails would indicate separate tallies independent of rejected/received/sent to me, i.e. "counting them separately". No?

[tonioo]

What I wanted to say is that you can choose to accept SPAM and move it to a Junk folder, quarantine it or drop it. In cases 1 and 2, the message is accepted so it will be counted as 'Received'. In case 3, the message will be counted as 'Rejected'. If we want a special treatment for spam and virus, then they should not be included in those counters but it looks like a incomplete information to me... What do you think?

…

On 10/05/2017 18:17, Der Mundschenk & Compagnie wrote: @tonioo <https://github.com/tonioo> That would seem to be a contradiction. The lack of information about what to do with |spam| and |virus| mails would indicate separate tallies independent of |rejected|/|received|/|sent| to me, i.e. "counting them separately". No? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#18 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAgc7oLpjPEdHw1quyjCi9iE60PudJxaks5r4eMKgaJpZM4NNa_q>.

[mundschenk-at]

I think it's impossible to handle this in a general way. We would need to parse configuration files for different milters (Amavis, Rmilter/Rspamd ...) or recreate the various possible settings via parameters. I'd like to show the difference, but that would also increase the number of graphs ("spam detected, but accepted" vs. "spam rejected") and I doubt we could do a generic and solid implementation.

Maybe it would be better to have spam and virus mails in a separate GraphicSet to show that it's on a distinct scale?

[tonioo]

Why not. Then we could implement specific drivers (amavis, rspamd, etc.)

[mundschenk-at]

@tonioo Basic Rmilter/Rspamd support has been added via #23. It should work exactly like the existing Amavis implementation (i.e. it doesn't change the way counters work currently).