Use the same port for TCP and WebSocket

[NihiliumOnMarz]

I am trying to listen on /tcp/1111, /tcp/1111/ws, and /tcp/1111/tls/ws and getting error bind: Only one usage of each socket address (protocol/network address/port) is normally permitted."}

I thought LibP2P was capable of this?

[Jorropo]

This has to be implemented manually in each transport (see https://grep.app/search?q=quicreuse&filter[repo][0]=libp2p/go-libp2p).
AFAIK this is only implemented across quic* and webrtc transports now.

[NihiliumOnMarz]

This has to be implemented manually in each transport (see https://grep.app/search?q=quicreuse&filter[repo][0]=libp2p/go-libp2p). AFAIK this is only implemented across quic* and webrtc transports now.

Looking through the code in TCP transport it has an option for reuse port but I do not see it in websocket transport. Is there an easy way to add?

[Stebalien]

Looking through the code in TCP transport it has an option for reuse port but I do not see it in websocket transport. Is there an easy way to add?

Not easily, no. It's not just a matter of saying "reuse this port", you need to be able to detect that the connection is not for you and pass it off to the correct transport.

[MarcoPolo]

@NihiliumOnMarz is there a specific use case you need this for?

[Jorropo]

Related libp2p/specs#608 I'll submit a PoC.

[aschmahmann]

is there a specific use case you need this for?

I'd like to have this in kubo. In particular, ipfs/kubo#10521 will allow us to automatically grant people WSS connections that are usable from browsers (see #1360) alongside other libp2p connection types (e.g. TCP+TLS/Noise + Yamux, or QUIC). Getting people to open up new ports with their docker configs, firewall rules, port forwarding, etc. is generally a pain and seems likely to hurt adoption. So if we can reuse the TCP port for WSS connections that would be pretty convenient.

@Jorropo my guess is that we can probably reuse a chunk of the investigative work (and perhaps some of the code) from the magiselect proposal here (thanks for that 🙏), and get some of the work landed even if the "look at magic bytes instead of using multistream select" proposal hasn't landed.