Unify websockets framing.

[majek]

@mrjoes on the mailing list http://groups.google.com/group/sockjs/browse_thread/thread/709ae905e315e1a6 noticed that once in the sockjs-protocol we do:

    ws.send(u'["a"]')
    self.assertEqual(ws.recv(), u'a["a"]')

and later:

    ws1.send(u'"a"')
    self.assertEqual(ws1.recv(), u'a["a"]')

[majek]

Bad news. Sockjs-client is using the form that isn't documented in the framing section of the protocol spec, ie: it just pushes raw string, without the parenthesis:

https://github.com/sockjs/sockjs-client/blob/67af34fc94df4fd08415c0d1aad6ad9dd2b048bc/lib/sockjs.js#L214
https://github.com/sockjs/sockjs-client/blob/67af34fc94df4fd08415c0d1aad6ad9dd2b048bc/lib/trans-websocket.js#L33

[majek]

I decided to always use squares - like ["a"] .

The disadvantage is two bytes more per message.

But it will give us a tiny bit more flexibility in the future. (I'm thinking about support for binary data already).

Hopefully this was long tested by sockjs-protocol so the core protocol doesn't need to change.

[majek]

Btw, the framing with square brackets was tested for a long time here:

sockjs-protocol/sockjs-protocol-dev.py

Line 550 in ae12ddf

ws.send(u'["a"]')

[fafhrd91]

i think all essential tests should be moved to latest version of protocol (WebsocketHybi10) or at least should be duplicated
in both hixie and hybi

[majek]

Agreed. Please file a bug if you have anything particular in mind.

[fafhrd91]

ah, i see sockjs-protocol-dev.py already sends message with square brackets

[majek]

Continued: #61