block confidence: Bad validator bot

[sandreim]

Short term

Create a bot that ingests information about backing points from the chain.

A good initial source for this data is Turboflakes. It seems to already have some push notifications:

Details

Once a day, the bot should ping the proper validator lounge channel and prompt the operators to fix it.

Long term

Gather direct contact information for validators. Potentially require validators to submit this information (or hash of it) on-chain.

Bonus: staking dashboard integration, so people can also see validator “outages” on top of points.

[bkchr]

IMO it would make much more sense to remove validators from the active set if they don't fulfill certain requirements. This means we track how many backing points they should have achieved, if they are not able to achieve 75% (random number) of them, we remove them. Removing means that you get "banned" from getting re-elected for X days and the operator also needs to actively put themselves up for "vote" again.

[burdges]

We should be resiliant against screw ups, especially if someone simply shiutsd own. It's a bigger problem if a validator no-shows lots but at least http://github.com/polkadot-fellows/RFCs/pull/119 should cut their rewards significantly. If they no-show but then catch themselves, and thus still get paid, then we could tweak that system so they do get paid less.

[Sajjon]

Here is the repo of the bot: https://github.com/turboflakes/one-t

[sandreim]

IMO it would make much more sense to remove validators from the active set if they don't fulfill certain requirements. This means we track how many backing points they should have achieved, if they are not able to achieve 75% (random number) of them, we remove them. Removing means that you get "banned" from getting re-elected for X days and the operator also needs to actively put themselves up for "vote" again.

Yes, this is also being considered: #5674 . But, I think we need to be careful, because somebody can just DoS validators and get them removed, especially if we use some higher threshold (75%).

The bot should be a softer first step towards this.