by Arra Derderian
14. May 2011 16:54
The other day I read an article on Mashable about a new JS library called Now.js. This article called Will Node Power the Next Twitter? NowJS Is Betting On It discussed how many believe the next phase of the web, Web 3.0, will be powered by servers running on libraries such as Node.js. I watched the video included in the article that depicted a user setting up a node server and then setting up a now.js chat room. It was a cool video that showed a new concept of having server and client both running the same code and managing the users connected to the chat.
NowJS Introduction: Simple Chat Server in 12 Lines of Code from NowJS on Vimeo.
After watching the video I was did some reading around the web at various sites trying to find the pros and cons of a JavaScript based web server. To me it seemed like the argument for node.js was that it is event based where traditional web servers are synchronous in nature. Node.js can handle many requests because they delegate the work and wait for a callback rather than waiting for the request to finish. This sounds like an optimal way to handle requests, but isn't it still only a good way to handle requests? Won't the server eventually be just handling requests so fast and pushing off the work that the system will slow down? Sure the initial request will get through quickly but the time of response will be longer as request amounts increase.
A couple other things I find it hard to believe is that all of the benchmarks done out there are indicating node.js blows away all of the traditional web server technologies. For starters, a compiled program should run quicker than an interpreted language. I know Google's V8 is lightning fast for a JS interpreter but it is just a basic concept I figure has to be brought up. Secondly, while node.js is putting out these great benchmark times it is important to remember it does not contain the modules that a Microsoft IIS has. Security, request filtering, error handling, and much more and not included. Writing out the headers of the document being returned is a manual step as shown here. I am still searching for some of the benchmarks and will post them as I find them.
Anyone who can comment on my questions here would be great. I would like to hear peoples thoughts. In the meantime I am going to try and download some of the pre-compiled libraries of node.js and get it going on Windows. Also I saw another cool walk through of implementing node.js in Windows Azure Cloud. Check them out with me and lets see what Web 3.0 is all about.
Subscribe