This post is more or less my notes from a moment of intense burning curiosity earlier in the day. I'll tag this if I come back to it but it was a compelling enough thought to stand up a CentOS 6.4 VM in VMWare Player, built node 0.10.26 and started installing some libraries out of pure curiosity.
Is there a reason why not to write a backend in Node.JS with WebSockets other than the obvious issues with performance, scalability, stability, availability, etc? And has anyone done it? Turns out the answer is kind-of sort-of.
And why would you want to? Except why not? We're doing a thought experiment and Google makes thought experiments easy.
From a quick look, my favorite so far is Game Closure. This has the most promise but maybe because it has clever front end web design and the curious hordes are easily sucked in by clever web design. It may have some horrible flaw I'm missing but it doesn't look it. This looks worth downloading and taking out for a spin. The open source license helps.
Back in time I spent some time thinking about writing RPG support tools for the iPad using several of XCode's provided data management libraries and a bit of Cocos2D for nice sprites, layout, and die rollers. Programming in Objective-C is a terrible experience but XCode tries to make the experience less of an anal probe. I made it through most of the Cocos2D tutorials before I had to give it up.
That's the front end. Interestingly I found two examples of backend -- BrowserQuest (code here) and Osmus (code here). I found a talk here. It's about a server framework called Pomelo - which is entirely in Chinese so your mileage may vary. Looking at the source code available in github is educational. All the examples do rely on Websocket management to drive connectivity between persistent web browsers and the backend.