Re: [hybi] #1: HTTP Compliance

[Jamie Lokier <[e-mail]>]

Ian Hickson wrote:
> On Mon, 17 May 2010, Dave Cridland wrote:
> > It makes sense that if we were to use WebSockets in a similar way, then 
> > a desktop client might initiate a session with an HTTP service without 
> > knowing a priori if it's an HTTP or WebSocket service.
> 
> I don't understand how that could happen. Could you elaborate on what kind 
> of desktop client you mean? When would you ever connect to a port without 
> knowing ahead of time what protocol you want to talk?

For example, a social networking client running in a browser, designed
to fetch feed information from numerous different services from other
providers (imagine Yahoo, Facebook, Twitter etc.), those services
running a variety of feed protocols over WebSocket.

When the user is given a URL for a third party bidirectional service,
the client tries WebSocket subprotocol A first, then WebSocket
subprotocol B, then C, then a HTTP hanging-GET equivalent.

Although you _could_ give the protocol along with the URL to users,
clearly that would be avoided as much as possible.  Less is better.

And anyway, if you did that, it would break when the provider "Yahoo"
switches their service from standardised subprotocol B to A (because
it works better in some way, and which has no visible effect on
Yahoo's own clients).  So the multi-service client needs to try them
in turn, not be told what to use.

-- Jamie