devp2p: discv4 & discv5

[decanus]

This issue is mainly here as the work process for the "Research: Discv5 feasibility" issue. There are several goals that should be achieved through it to reach the final study. This includes exploration and further understanding of the topic & technologies relevant.

Discv4 & Discv5

Firstly, the technical details of discv4 & discv5 should be analyzed. How doe each of them function, what guarantees do they provide and where do they fall short? Then additionally the difference between the 2 should be highlighted.

Next, Felix Lange's pull request ethereum/go-ethereum#20750 should be analyzed to see how and if it differs from the current spec.

Finally, we analyze the difference between the current, ethereum/go-ethereum implementation and that within nim-eth. Where do they differ, what needs to be added etc.

All together this will provide enough knowledge on the subject.

It may make sense to later look at discv5 and investigate the changes that are still to come and think about changes we could still get into the protocol that would help improve the overall functionality.

Literature

• Node Discovery Protocol - post
• Node Discovery Protocol v5 - post (discv5 unlike v4 is made up of various documents)
• Discovery v5 MVP - issue

[decanus]

discv4

discv4 is a Kademlia like DHT, this structure was chosen as it provides an efficient way to organise nodes has a topology of a low diameter.

secp256k1 keys are used as Node IDs the distance is calculated using the XOR of the hashed public key.

Nodes are expected to maintain up to date ENRs, they can be requested by other nodes using the ENRRequest packet.

[decanus]

IMPORTANT: discv5 is currently still in DRAFT and could make breaking changes at any point.

[decanus]

Weird thing in discv5 is that you need to call REGTOPIC even though topics could be included in the ENR, see issue here: ethereum/devp2p#144

[decanus]

A full summary of the research compiled during this issue can be found on vac