#yggdrasil

BTW, вот эта фича, которая будет в новой версии
The routing algorithm now tries to minimise both the cost and remaining tree distance to the destination, which should improve some cases where direct paths were ignored in favour of indirect paths
Меня не то что-бы радует. Потому что у меня есть живая ситуация, когда трафик идущий по пути A<->B<->C идёт быстрее, чем когда он идёт A<->C. И вот, дпустим, #mycelium это отрабатывает. а #yggdrasil и раньше - не всегда такое ловил, а сейчас, как я понимаю,  вобще такое ловить перестанет.

Печалька.
twinkle.lol

@ufm We're planning to rethink link costing in v0.6 so that we have better multi-hop awareness of latencies rather than just next-hop awareness.

At the moment what we do is optimistic at best, but in the worst cases it can pick a very fast next-hop but then the rest of the onward path is worse than the direct peer would have been. Then people complain that the direct peering shouldn't have been ignored.

In an ideal world, people would fix their peerings too. In reality, that doesn't happen.

@Neil
You know, I'm a consistent supporter of being able to specify a link "cost" instead of dropping a peering just because it's bad. It's better to keep a very "bad" (or very "expensive in terms of money", for example over mobile data) link as a fallback than to have none at all. But anyway, I'm looking forward to the new version with interest - I have a place to test it. :)

And good luck to you!