Re: [hybi] Multiplexing extension spec draft 03

[John Tamplin <[e-mail]>]

On Wed, Feb 29, 2012 at 5:03 PM, Greg Wilkins <[e-mail]> wrote:

> Sorry, but I'm not understanding either choice you are implying exists.
> Can you expand with examples?
>
> Masking is a per frame operation, so I don't understand how it can be
> applied to a logical channel?
>

You're right -- that was left over from when masking had a per-connection
key and maintained state.


> True, but the cost is passing on extra complexity and work to the server
> and tools.  One of the purposes of having intermediaries is to offload work
> from the servers, so if they can do the extra work to flatten multiple
> multiplexed layers into a single channel space, then that will be of great
> assistance to the servers.   Requiring muxing multiplexers to demux any
> incoming muxed streams feels like a lot less work and complexity rather
> than requiring all servers and tools to support an arbitrary depth of
> nesting. (So says the server developer....   shat say the intermediary
> developers?).
>

I think of a router or switch where you want to minimize the processing as
you pass bits from an input to an output.  A WS mux intermediary is likely
to be higher level than that, probably about the same as a load balancer so
it doesn't necessarily apply at that level.  I can see the argument both
ways, and I think we need to discuss the relative merits of both before
deciding.

-- 
John A. Tamplin
Software Engineer (GWT), Google