Evaluate WebTransport protocols and APIs

[endel]

There's a draft on WebTransport going on, which is a successor of WebSocket that supports both reliable and unreliable communication. It's not currently usable - hopefully, it will in the future (when?)

It would be nice to start thinking about how to make Colyseus use its reliable/unreliable API for sending and/or receiving data.

[endel]

According to w3c/webtransport#65 (comment), we could use the RTCQuicTransport as a proof of concept, before WebTransport is actually implemented.

I honestly don't think WebRTC is viable for authoritative servers, but we could use RTCQuicTransport's communication layer to be prepared for supporting unreliable messages while WebTransport is being developed and matured.

[ouya99]

why WebRTC not viable for authoritative servers?

http://geckos.io/

[yandeu]

why WebRTC not viable for authoritative servers?

It is! It was initially just not meant to be used on the server. But there are now WebRTC libraries for Nodejs, Golang, C++, Rust, Python and surly many more.

[IfnotFr]

Just sharing something for the brainstorming. we were talking about it with ECSY users. The http://geckos.io/ guy is showing some real improvements with bad connections.

200ms and 10% packet loss TCP vs UDP :
https://youtu.be/ZEEBsq3eQmg?t=104

This is impressive !

[yandeu]

200ms and 10% packet loss TCP vs UDP

To be precise, the latency to the Package Loss Test was 225ms (Georgia US). The latency to the EC2 server was 130ms (Frankfurt DE). If the latency was >200ms, WebSockets would have performed much worse.

[IfnotFr]

Oh you are this guy, I did not noticed, i am ambarased :D

... Thanks for the precisions btw, your tests shows why UDP is important :)

[GameDevBhavesh]

@endel any plan about implementing http://geckos.io udp socket for colyseus ????

[jerrygreen]

I wanted to find some handy ways to do networking in haxe/heaps game, and this looks quite good, and potentially just excellent with UDP... Looks real promising, especially with that proof of concept video posted by @IfnotFr (originally by @yandeu).

Now I wonder what new API it will bring into the library, because of the nature of UDP, it's only beneficial in some particular cases, like when you have little pieces of data that you want to update as frequent as possible, like sending position of a unit, or position of a mouse, etc. Any ideas?

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 7 days.

[github-actions]

This issue was closed because it has been stalled for 7 days with no activity.

[epurban]

Hi @yandeu! I was wondering if you and your team plan to add WebTransport support once webkit implements it. They have accepted it to their standard positions. We would love to use the benefits of UDP through Colyseus for our multiplayer game! Thanks and have a great weekend :)

[endel]

Hi @epurban, thanks for your interest here!

I've been experimenting with the fails-components/webtransport implementation of WebTransport in the webtransport branch. The JavaScript SDK also has a webtransport branch and things are mostly working, though there are still a few things to be improved/refactored to unify the transport API's.

So far, in addition to the send() method, I've added sendUnreliable() both on the server and client. By default, both a bidirectional stream and the "datagram" stream are being read for every connection - this might not be a good idea for every use-case...

The fails-components/webtransport hasn't been used extensively, so it might have some limitations, it would be really valuable if we could get more people involved in experimenting and helping test out.

(I'm reopening this issue to continue the discussions...)

[freeqaz]

Thanks for the update, @endel! @epurban and I will try slapping that in place for our existing Colyseus app to give it a spin.

If we find any bugs should we open bugs as new issues? Or just document them in this thread? Happy to help however we can!