On Thu, 22 Jul 2010, Simon Pieters wrote: > > Because I expected the closing frame to be the exact sequence 0xFF 0x00 > and nothing else. It makes the protocol simpler to understand and > explain.
This is a moot point now. > The problem I was discussing is the following scenario: > > 1. Client opens a websocket without any specified subprotocol. > 2. Server replies with subprotocol 'räksmörgås'. > 3. Client accepts the connection. > > Next time, the client cannot open a websocket with the subprotocol that > the server used, since räksmörgås is non-ascii and will raise an > exception if used in the constructor. Hence, I think the server should > only be allowed to use ascii subprotocols (without spaces) to conform to > the spec. Fixed. -- Ian Hickson U+1047E )\._.,--....,'``. fL http://ln.hixie.ch/ U+263A /, _.. \ _\ ;`._ ,. Things that are impossible just take longer. `._.-(,_..'--(,_..'`-.;.'