[nkzawa]

Add the permessage-deflate extension support for WebSocket, and http compression for polling. They are enabled by default.
Also you can disable the compression per message through the option of send method.

Related: socketio/socket.io#1148, #298 #271

[nkzawa]

It seems we should have a minimum size threshold for http compression, since compressing small data makes them larger.

[leearmstrong]

When will this make it into the live SocketIO code? Anyway to use it now?

[defunctzombie]

@nkzawa Why wouldn't we just let the proxy handle this?

[nkzawa]

@defunctzombie mainly because we want to turn the compression on/off per message(payload) to emulate the permessage-deflate extension of WebSocket. I think we can't support this with proxy.
And also we can't use Content-Type header to decide whether to compress data or not like a http server.

[3rd-Eden]

@defunctzombie you can still have your proxy handle this by turning compression off in engine.io right?

[nkzawa]

@lpinca updated, thx!

[rauchg]

typo

[leearmstrong]

Any ETA when this will be merged. This could be huge!!

[nkzawa]

@Leeus I'm working to reflect code review. Maybe this will be merged after that if there is no problem anymore.

[nkzawa]

@rauchg fixed and rebased. could you check again?

[rauchg]

When we call sendPacket manually, should the compress flag be undefined so that it obtains it from the default provided by the configuration in the constructor?

Otherwise, it seems that if we disable compression globally, it'll still force to true when we send non-user packets (like noop)

[rauchg]

Before, it seems that it was always defaulting to compress: true if it was undefined?

[nkzawa]

compress: true for sendPacket(send) is just one of the conditions, so it's ignored when compression is disabled globally and when client doesn't support compression.

[rauchg]

@nkzawa please rebase so I can merge.

[nkzawa]

@rauchg rebased