Matthew Toseland
Matthew Toseland
I thought the opennet-changes branch did this?
So in conclusion, this is longer term ... #491 should be considered for 1472...
Maybe. I'd want to simulate it first, and have much better stats gathering (digger3's stats are down). Bootstrapping would be dramatically slower because new nodes would get no nodes from...
We could test this without a full blown opennet simulation. The necessary changes will work fine on their own on the end-node. E.g. bootstrap a new node with this branch,...
Bear in mind that there is IP-address-based limiting on the seednodes, and announcement is rather hit-and-miss anyway. So bootstrapping more than one node on the same IP in the same...
Also, this patch should have _no_ effect on announcements - except to make new nodes more dependent on them.
Ok so you've confirmed that it works but it takes longer to bootstrap, which is pretty much what we expected??? Re hops taken, you can't deduce much from success rates......
Also, we spend 2 hops at HTL 18 on average.
It would be interesting trying this out with my simulator. Results so far suggest that the packet scheduler layer is very costly. However I haven't reviewed it yet, sorry...
I've added a bunch more fixes, I think this is reasonable now. Please re-review. :)