[falcolas]

No, it's really not that easy, sorry.

Looking at the code, you're not actually doing anything with websockets here. There's no handshake parsing, there's no open socket object to pass data back and forth on, there's nothing actually websockety about this.

Instead, you seem to have a simple ajax endpoint.

Or am I missing something?

[EDIT] Clarifying why I said it's not that easy, since that seems snarky as I read it again. To properly do websockets in Django, you have to write your own runserver, which allows processes up the stack to capture the raw socket, perform the websocket handshake methods, and do proper parsing of the incoming data to break the messages on the sentinel values.

I recently tried to do this, and the one package that did this wouldn't work with Django 1.6, and we would have to give up our flup runfcgi, which I wasn't ready to do.

[gizmo]

I only looked at it for a second or so, but it looks like it's a simple json proxy. You put a webserver with persistent connections on one end (Tornado in the example code) and a regular django environment side by side. The web socket requests are translated in to GET/PUT/POST requests, are pushed to django, django responds and the response is pushed to the client.

So from django's point of view there are stateless JSON requests and responses, so django's happy. The python proxy code is a lightweight wrapper that doesn't have to understand any business logic, so that's pretty straightforward as well.

[falcolas]

I see where I was confused; I looked at the actual code itself, not the example where tornado was living and being used.

What are the advantages of this over normal AJAX requests, given that it's all using the single request, single response mode of operation (and you can compress/cache AJAX requests/responses, which websockets doesn't support yet).

[mcantelon]

Seems like advantage is skipping network roundtrip to/from client browser. But still seems expensive in terms of cycles.