feat: add kcp transport with listener implement

[derrandz]

Description

This PR aims to introduce KCP as a supported protocol in the libp2p library

Missing

• Metrics
• Reuseport

[marten-seemann]

Which issue does this resolve? I’m not convinced KCP is a good idea, but we should discuss this in an issue and not on a PR.

[marten-seemann]

There already is a 3rd party KCP transport implementation: https://github.com/paralin/go-libp2p-kcp.

[marten-seemann]

#193 was the tracking issue, and we closed it once QUIC was merged. There really shouldn’t be any reason to use KCP if you can use QUIC instead.

[derrandz]

Hi @marten-seemann thanks for the context. This was old and wanted to do a performance test against quic through libp2p to provide a rational for using kcp alongside quic or instead of quic. I would have to come with rationale and benchmark stats but closing this for now makes sense.

Cheers,

[marten-seemann]

Hi @derrandz! I'd be very interested in seeing the performance numbers :)

Note that libp2p is designed to allow you to use 3rd party transports without getting them merged into go-libp2p, as long as they implement the transport interface. All you need to do is pass them to the libp2p.Transport constructor.