kad: emit ToSwarm::NewExternalAddrOfPeer

[thomaseizinger]

Description

With #4371 merged, we have the foundation of #4302 implemented.

libp2p-kad needs to be extended to emit this new event whenever it discovers a new address through the DHT.

Motivation

Other behaviours should be able to learn the addresses discovered by kademlia.

Current Implementation

Only kademlia knows about its discovered addresses.

Are you planning to do it yourself in a pull request ?

No

[1010adigupta]

would like to work on this

[1010adigupta]

would work on mDNS and DHT simultaneously, you can assign me this as well

[thomaseizinger]

would work on mDNS and DHT simultaneously, you can assign me this as well

Make them two PRs please! :)

[justcode740]

unsure whether this one is taken but seem prev one has made such claim and not submit pattern in some other repo couple times..

[thomaseizinger]

You are welcome to work on this too!

[blacktoast]

Hi, do you mind if I try to do this?

I know a little bit of Rust, but this is my first time contributing to a p2p related library, so if you don't mind, is there any code or documentation that would help me solve this issue?

[dariusc93]

@blacktoast are you still interested in tackling this issue?

[PanGan21]

Hi guys, I saw that this ticket was stale for some time so I gave it a try at #5549 . Let me know what you think 🙂

[guillaumemichel]

Following up the discussion from the merged PR #5549.

With #5549 a NewExternalAddrOfPeer event is only emitted when a new peer is added to the routing table, and not when a new peer is discovered.

I agree with @Wiezzel that it would be best to emit the event when the address is discovered, so that an event is emitted for all (newly) discovered peers.

Though as we don't have a peerstore, it means that there may be many events emitted for the same node (and same address) every time the node is included a kad response. We could keep a state per query, which would already limit the number of duplicates, and not generate an event for peers that are already in the routing table, but even then these event will be noisy.

[dariusc93]

I agree with @Wiezzel that it would be best to emit the event when the address is discovered, so that an event is emitted for all (newly) discovered peers.

I kind of agree that we should probably emit on newly discovered external addresses and not just when adding to the routing table (which is the purpose of the event anyway).

[PanGan21]

Agreed. I will check what is the best place to move this, so I can do a fix pr if you agree. Any pointers are welcome!

[drHuangMHT]

Is this done? #5549 has been merged.

[elenaf9]

Is this done? #5549 has been merged.

👍 yes, thanks.

[dariusc93]

Is this done? #5549 has been merged.

Is it but it looks like discussion was going to continue to determine if we should emit only when adding to the kbucket or upon discovery of any peers regardless of their position in the kbucket. Not sure that been resolved, has it?

[elenaf9]

Is it but it looks like discussion was going to continue to determine if we should emit only when adding to the kbucket or upon discovery of any peers regardless of their position in the kbucket. Not sure that been resolved, has it?

You're right, I should have scrolled down more on that PR 😄. Still, I would consider the current issue to be resolved, kademlia does emit a NewExternalAddrOfPeer event when discovering new addresses. If the current implementation of it should be changed / adapted to a specific use-case, a new issue should be opened IMO.