Support connecting through proxies about emacs-websocket

[catern]

I got some extremely hacky modifications for this working, where, when a proxy needs to be used, I synchronously connect to the proxy, send a CONNECT, call gnutls-negotiate, then proceed with websocket.el's normal processing. I'll post the patch here in a few days once I feel a little more comfortable with it. But for now it seems to work.

[ahyatt]

I'll have to think about this - it seems like a good thing to have, but it'd be nice to do this in a way that wouldn't complicate the main websocket code. Let me know when you have a patch, and I'll take a look.

[catern]

Here's the patch: #74

I did a lot of digging into url.el's proxy support, and was able to reuse a fair bit of functionality - the main thing that I couldn't reuse is url-https-proxy-after-change-function, but maybe there's a way... it would be nice if url had a way to say, "establish the connection, including proxy setup, then send the request, then just call my process filter and get out of the way".

[ahyatt]

Interesting. I don't think this is that hacky, however I'll have to think about this.

Also, do you have signed FSF papers? This probably is small enough that it doesn't matter, but I need to verify this if you don't.

[catern]

Well, the main hacky bit is the sleep-for. Everything else is fine I suppose, but relies on that core hackiness...

I don't have signed FSF papers, but I've been meaning to for a while, so I'll go ahead and do that now, regardless of whether this is small enough or not.

[dickmao]

How does failure manifest itself? It would seem from catern's changes that websocket-open simply can't connect, and that some kind of initial negotiation with the proxy is required before opening the network stream to the intended target. I have very little experience with proxies -- I barely know what they're for other than to make our lives more complicated. :)

In regard to millejoh/emacs-ipython-notebook#754 , the websocket from the emacs client does get through, it's just the remote server doesn't see an auth token in the json message. That is quite a mystery, but at first glance this PR wouldn't remedy that, or would it? A proxy presumably would not strip data from a message.

[ahyatt]

Yes, I think this sounds like a different problem. This problem should result in connection issues, not authentication issues. That said, I haven't tried replicating this - I don't have a proxy, or a websocket server that does auth checks using json. And I'm also not that knowledgeable about proxies.