No support for cookies in websocket header about emacs-websocket

[ahyatt]

Thanks, let me look into this.

[ahyatt]

Looking at this more, I think the only reasonable way to support cookies is to tightly integrate with url-cookie. If we support sending cookies, we probably should support handling set-cookie as well. That means the client needs to call url-cookie-store before the call to websocket-open. I'll proceed with this kind of implementation unless you have some specific objection.

[ahyatt]

Also, about the port issue - your code implies that the port is needed even if the port is unspecified (in other words, port 80). If true, it seems like a violation of the RFC, which says:

1. The request MUST contain a |Host| header field whose value
   contains /host/ plus optionally ":" followed by /port/ (when not
   using the default port).

[ahyatt]

I've written a fix for the port issue to commit b7a1907. I've followed the RFC, though - so no port if it is the default.

I'll merge this to the main branch after you OK it.

[millejoh]

Tested b7a1907 with ein and looks okay to me. EIN was working fine prior to this change, but I discovered not specifying the port in the request header was causing issues when dealing with secured IPython notebook servers. The problem is that without the port Tornado looks in the wrong place for the cookie. Not the most sterling technical explanation, I know, but I hope it makes some sense.

Also, I agree with the approach of calling url-cookie-store. Thanks for the help!

[ahyatt]

Please take a look at 542c6fe and tell me if it works for you. All tests pass, but I don't have an real server to test against with actual cookies.

[millejoh]

Works for me. Thanks very much for implementing this!

[ahyatt]

Both changes are now merged into the master branch, and I'll wait a few weeks to see if anyone reports any problems, and release a new version thereafter.

If you'd like me to release before that to fix the user problem, that's fine.