Some problems with this approach: 1. Concurrency. Regular HTTP and pipelining allow the transfer of more than one resource at a time, often times many more. 2.

Some problems with this approach: 1. Concurrency. Regular HTTP and pipelining allow the transfer of more than one resource at a time, often times many more. 2. Caching. The author mentions this, but fails to mention browser-side caching, which is the most significant form of caching, that is, to get a resource without having to go out over the network. This could perhaps be addressed with local storage, but his node module doesn't take that into consideration. 3. Compression. Regular HTTP requests support gzip; I'm not sure if websockets do or not. An initial google wasn't promising and it seems to be experimental. He mentioned in-browser unzipping, which is interesting but a more standards based approach would probably be via the content-accept header on a websocket connection.