Re: [hybi] WS framing alternative

[Ian Hickson <[e-mail]>]

On Fri, 30 Oct 2009, "Martin J. Dürst" wrote:
> > 
> >Actually the WebSocket-Protocol header lets you negotiate that at 
> >connection time.
> 
> Does that mean that WS allows arbitrary headers in the upgrade 
> request/response? Or that there is one header, WebSocket-Protocol, with 
> potentially a lot of data stuffed into it?

You can set the desired value for -Protocol header by passing a parameter 
to the WebSocket constructor. That sets the value from the client. The 
server then responds with the protocol _it_ supports. If the two match, 
the client provides the connection to the script. To support multiple 
protocols, the server can use the client-provided value to decide which 
one to use.


> You seem to assume that Web Sockets communication is basically just 
> between a client and a server, without any intermediaries getting 
> involved in any significant way.

Without any independent intermediaries, anyway. On the server side, there 
can be whatever intermediaries the server-side developer likes; they can 
all be considered to be the server-side implementation from the 
perspective of the client (and the spec).


> But if there's anything where intermediaries get involved in a specific 
> way (other than just letting everything through or just blocking), then 
> there's probably a need to let these intermediaries know some info. 
> Because the payload frames are for data exchange between the client and 
> the server, they are essentially opaque to intermediaries, so you can't 
> use them for that purpose. That's where a separate frame type comes in 
> handy.

What's the use case for independent intermediaries?

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'