Add sub-protocol support

[rstoyanchev]

As a WebSocket emulation layer, it makes sense to support sub-protocols in SockJS. The sockjs-client could have a "protocols" option added with the value passed to the server either through the W3C WebSocket constructor or by setting the Sec-WebSocket-Protocol header.

Servers can then choose to add support by checking the Sec-WebSocket-Protocol header.

[matzew]

👍 yes, that would be a nice addition

[sshcherbakov]

+1!

When will this issue be fixed?

[radicans]

Once people stop commenting and start contributing.

[sshcherbakov]

There are 7 open PRs, some of them hanging since 2013.
Seeing this picture one might ask whether the project is dead.

[radicans]

@sshcherbakov Yes, lets completely ignore the fact that @brycekahle has been had at work rewriting the client since he took over the project and posted various of updates on issues stating that a new release will come once the refactor is done.

The health of a project cannot be determined by the amount of open pull requests and issues. With that same mindset you would also consider joyent/node to be dead as it has 216 open pull requests and 710 issues. Where the oldest pull request is from Mar 10, 2012.

[sshcherbakov]

Sorry, didn't check the recent closed PR activity.
It looks definitely better than half a year ago.

[brycekahle]

Since this needs to support more transports than just Websockets, we can't use the WS specific headers. I think the approach taken in #139 is better, by adding it to the info response.

I'm going to schedule this for 1.1 given how much work it will probably be.

[rstoyanchev]

This would be a good fit for 1.0 IMO especially now that #65 is addressed. SockJS is very much focused on being a transport so it makes sense to have some sub-protocol negotiation support.

[danbev]

@brycekahle Let me know if there is anything I can help out with on this issue.

[brycekahle]

This would require server work and changes to the protocol. Right now 1.0 is compatible with the existing protocol, so I don't want to delay 1.0 even further with this work.