Re: [hybi] Apples and Orangutans

[Jamie Lokier <[e-mail]>]

Greg Wilkins wrote:
> >I don't understand why a 16bit integer solves the problem for TCP, but a 
> >string doesn't solve the problem for WebSocket. They seem exactly 
> >equivalent to me -- they identify the target of the connection, and by 
> >convention, the protocol that that target supports.
> 
> The big difference is that for the web you only have 2 strings
> while TCP/IP has an infinite number of 16 bit integers.
> 
> ie The TCP/IP model is that if you want a ssh and 2 ftp conversations
> then you open one connection to port 22 and 2 connections to port 21

That's separate connections, to the same service, not to separate
services.

If want to connect to different services on the same host, over TCP/IP
you typically need something like separate IP addresses (bad idea) or
a distinguishing "virtual host" header like HTTP's Host:.  Or
different ports, but then it's not a well known service and often
doesn't work over firewalls.

The analogous thing in WebSockets would be a path for the resource
being connected to, and a string/path to identify the service type.

However, I don't think there is really a need for the second string.

The first one is enough to identify both the resource being connected
to and the service type, in the usual ways that people assign
application-specific meanings to URLs.

-- Jamie