Re: [hybi] updated Charter proposal (WebSocket)

[Mark Nottingham <mnot at mnot.net>]

On 27/10/2009, at 12:24 PM, Ian Hickson wrote:
> I would expect server-side implementations to translate incoming  
> WebSocket
> connections into appropriate internal connections; e.g. if Example  
> Corp
> uses RPC extensively internally in their datacenters, a front-line  
> server
> could receive a long-lived WebSocket connection (speaking that site's
> subprotocol) and translate it into short-lived RPC on the backend.
> However, that would be layered above WebSocket just like this kind of
> thing is layered above hanging-GET systems today.

Yeah. In theory it would be nice if, for example, a corporate or ISP  
proxy could "manage" a bunch of WS connections back to a third-party  
service provider, but there's a lot of risk and overhead to going down  
that path, and not a tremendous amount of benefit.

I suspect most of the intermediary cases that are going to come up for  
WS in the near future are going to be server-side (like HTTP  
accelerators/reverse proxies), and they're already under the control  
of the server, so they can coordinate.

--
Mark Nottingham     http://www.mnot.net/