Re: Shared Locator Address Pool (SLAP) protocol proposal

[Dave Crocker <dhc@dcrocker.net>]

Folks,

Further thought:

SD>> stopped. But we've clearly considered providing a common capability
SD>> for all transports in a situation that was architecturually similar,
SD>> and the ECM pictures had congestion management in a shim underneath
SD>> multiple transports (the same point in the stack, if I get the SLAP
SD>> vision).

DC> I think it is the same point in the stack, yes. In fact, the discussions
DC> that led to my suggestion using "connection id" as the fine-grained
DC> identifier for a pool also prompted me to realize that the new, IP
DC> Endpoint (IP-EP) module also needs to provide common congestion control,
DC> as I commented in my parallel proposal to SLAP.

With a multiaddressing pool mechanism, the transport may not see the
locator information, instead seeing only an identifier.

However, differential congestion control is needed at the locator level.

So I think the long-term effect of creating pools needs to be to bring
congestion control into the core of the pool mechanism.

Think of this as taking the bottom end of transport and the top end
(sort of) of IP and making them overlap.

Mumble.


d/
--
 Dave Crocker <dcrocker-at-brandenburg-dot-com>
 Brandenburg InternetWorking <www.brandenburg.com>
 Sunnyvale, CA  USA <tel:+1.408.246.8253>