Unban lists - followable rooms with ban overrides

[bkil]

If I don't agree with certain bans on a foreign list, provide a way through which I can unban that given person on our mjolnir instance.

This should ideally be done through "unban lists" - rooms where overrides are being published that other mjolnir instances can also follow, similar to how "ban lists" are implemented.

Related: #228

[Gnuxie]

I'm going to put my own thoughts on this problem here. There is a problem in that you don't just want to unban someone. It's that you are disagree in list that ban is coming from in some edge cases. This is why I proposed the reinforcement part of https://github.com/matrix-org/matrix-spec-proposals/pull/3849 (The MSC in general needs a lot of work). The end goal being that policies you are likely to disagree with will require confirmation before your mjolnir blindly applies them.

[bkil]

Let me share the exact scenario. For day to day use, we would actually desire to follow the matrix-coc list because they add certain local bans an advance (while we sleep) due to having insider information.

Requiring confirmation would actually make following that list completely useless for us, because after some hours, our detectors, reporters and mods will have dealt with the abuse autonomously.

It does happen however (let's say in 1% of the cases) that the matrix-coc admin was heavy handed and banned someone without consideration, without discussing with other mods or they just refuse to appeal the ban later on after mitigating and educating the user. In such cases, it would be an acceptable (if not an ideal) compromise for us if mjolnir banned that user on sight and after we get up, we add that user to our own unban list. Then mjolnir could also do its dance with unban true on the user and perhaps reinvite her to the rooms where they have been banned.

[Gnuxie]

Well the idea is you wouldn't need to manually approve each policy in the list, only ones that conflict with the reinforcements you have made previously.

[bkil]

I see, so it would be basically #240 then that we have considered for our own ban tool, but that's a pretty hard nut to crack.

Well, I don't see how realistic it would be if I had to vow for the thousands of users within our community one by one and then do the same every day for the dozens of new users. But only after they talk and only after the first time, so I had to keep track of who said what where and how many times then.

At present, our new members are non-abusers in the great majority of time, and hence it would make it much more economical if you could default on this. For everything else, you could also easily keep track of the seniority of a given member automatically based on stats (e.g., active days, reactions, etc within our trusted rooms). So individual manual vowing for most users wouldn't be needed.