GitLab

dht doesn't need the whole network interface, only needs a Dialer.
(much reduced surface of possible errors)

Changed lots of peer use, and changed the peerstore to ensure
there is only ever one peer in use.

Fixed #174

![](http://m.memegen.com/77n7dk.jpg)

need to switch away from timing tests. its very annoying.
Should use proper go sync, as in
https://www.youtube.com/watch?v=ndmB0bj7eyw

for some reason travis has more goroutines running by def.

important to always reread your code.

It's a patch, really. it's not the full multiconn fix.

- does end properly
- no goroutines leaked!

This commit actually removes the previously introduced
chan net.NetMessage, in favor of raw []byte. It plays
nicer with crypto/spipe, and it makes more sense in the
context of a "single connection", i.e. I already know the
peer I'm talking to, from the connection. The NetMessage
additional Peer is useful swarm and up.

This Commit changes the relationship between Conn and Swarm.
After this, Conn is significantly more autonomous, and follows
an interface.

From here, it will be very easy to make the MultiConn (that
handles multiple Conns per peer).