Node Knockout Lessons
I participated in the Node Knockout together with Rafael Valverde and Claus Wahlers, and it was buckets of fun. We originally intended to do an application that involved taking page screenshots, but we were still facing problems getting it to run after a day into the competition, so we started over with a different idea and ended up delivering a functional application a few hours past the deadline, but it was still worth the experience.
NodeVote is a simple real-time, anonymous, voting room application. Like many people, we used Express.js, which is built on top of Connect (WSGI for Node.js). But in reality, the web framework was a last-minute decision and we went along with what seemed the most approachable. Express.js is very Flask and Sinatra-like, which was instantly welcoming. We also looked at Nitrode, but that looks like an attempt to provide a wider set of HTTP features (especially caching), and I'm not convinced that would be better than just using nginx and having it forward certain requests to Node.js backends.
What is clear from this experience is that Node.js is here to stay. Ryan Dahl is doing an excellent job evolving the project, which has very easy to follow documentation and a rapidly growing community. I'd encourage you to go and watch Ryah Dahl's Google Tech Talk on Node.js. It goes into great detail explaining why Node.js as an application server option is conceptually sane, and well, surprisingly fast too. He also starts the video with a great piece of advice, with pretty much sums up my feelings towards web application development these days (paraphrased, emphasis mine):
"There's a lot of different ways to write servers, especially for high-level languages, and everybody kind of employs their own flavor of how to abstract this problem of dealing with possibly thousands of different people connecting to your server at a time. And I think we need to be careful how we make these abstractions. The problem is that at some point you might have to use those and you just might feel trapped by them."
Rails, for instance, is a great example of one big set of abstractions that kept growing too big. Granted, it abstracts applications on top of a well defined architecture, but what I myself and I believe many other people realized after a while is that you don't need MVC all the time. Rails already allows you to use raw Rack HTTP handlers, and it's usually recommended that you do so for high-traffic resources. But what I believe, in fact, is that you can use these minimalistic HTTP handlers to build even the largest applications, while also experiencing greater ease in managing and evolving the code in the long run.
The more you understand the underlying protocols, and I mean everything from BSD sockets to HTTP, the less you'll want your frameworks to abstract away from you. You'll know how to choose what to use and not to use from them, and I think the idea here is that as long as your framework limits itself to simply exposing these underlying protocols you're using in efficient language constructs, you'll end up with incredibly succint and powerful application code.
I think that is what Node.js is about and that's why it's here to stay.