Better Connection Tagging

[aschmahmann]

We currently tag peers in our buckets higher if they are closer to us in Kad space. This is a bit unfortunate since we really want to tag peers that are useful to us (i.e. allow us to complete queries quickly and correctly).

Some aspects of this are:

1. We may actually want to tag buckets higher if they are further from us in Kad space as we are more likely to be performing queries against those nodes
2. We may want to have some peers in each bucket that we're connected to, instead of all from a few buckets
3. We may want to take some metric of query usefulness (e.g. whatever metric we use for routing table pruning) into account for tagging

Thoughts @Stebalien @petar @aarshkshah1992?

[Stebalien]

I think all of these are good ideas. We should also consider flat-out protecting some nodes. For example, we could protect the closest 20 peers and maybe the "best" 20 peers?

[aarshkshah1992]

@Stebalien @aschmahmann I've always wanted to get well acquainted with how connection tagging works and why we do it. This looks like a good entry point. Do you guys mind if I pick this up ? I completely understand what we are trying to accomplish here.

[Stebalien]

We have fixed this.