Replace Socket.IO with a lighter-weight alternative

[jniles]

This has been mentioned before, but I'd like to get a public poll. Socket.IO is a bit overkill for the current project scope. My vote is for the ws. In particular, look at the comparisons between frameworks here and here.

Has anyone had experience with other frameworks more suitable to agar.io-clone? Can we decide on a new alternative to put together some migration code?

[JoeMattie]

It needs to be done, no question about it. This is probably the most pressing issue preventing a game like this from scaling (with implementation of quadtrees or similar partitioning of players and entities being a close second).

I found some good info on the subject here:

http://buildnewgames.com/optimizing-websockets-bandwidth/

[c-hew]

I agree that WS is a better option. It solves many of the laggy issues!

[keon]

how about sock.js? I am planning on change to sock.js on my fork

[jniles]

@KeonKim , did you give sock.js a try? My concern is that sockjs doesn't use raw websockets and if we are going to the trouble of re-writing, we should be the lightest and best possible.

[keithagroves]

The Agar.io server implementation Ogar is using ws. Would that work well enough?

[igorantun]

I think so. ws is great.

[keon]

@jniles I haven't touched on it yet, I also think going for ws is better choice though.

[jniles]

Sweet. I propose that refactoring efforts focus on making sure we have testable components and test cases for those components, then we can swap out bits and see how it goes. It seems the consensus is for ws, so when the time comes we'll use that library.

[abalabahaha]

I seem to have gotten a socket.io -> einaros/ws transition on one of my branches, still needs a lot of testing though.

On a side note, every website that uses the current tagline would be wrong if we were to switch off socket.io.