On Fri, Jun 11, 2010 at 9:26 PM, Tony Li <[email protected]> wrote:
> Again, function 1 does not make the namespace topologically sensitive.  It's
> just using a flat identifier space and explicitly learning the location of
> the identifier.  This is clear if you consider the situation before the host
> generates a packet: before the switch learns of the identifier, it has NO
> idea where the host is in the network.

So it defaults to repeating the packet to all ports where a router
would have chosen to drop the packet.

That's the difference between an identifier and a locator? An
identifier defaults to flood while a locator defaults to drop?


> Function 2 is filtering based on an identifier.

Is it now? Last I heard it, different process subscribe to different
multicast MACs and can even subscribe to different unicast MACs (e.g.
a host OS with guest virtual servers). There's a decision about which
process to repeat the packet to. Just like a router, it defaults to
not repeating the packet if the destination MAC matches none of the
known destinations.


> Function 3 is simply including a source identifier.  It's only real purpose
> is to support the learning function for function 1.

Actually, I was talking about the destination MAC that gets tacked on
from the ARP table, not the implicit source announcement (much like a
route announcement) that lets the switch know which port to repeat
subsequent packets to.

But hey, you say poTAYto, I say poTAHto and we all dance a little jig.

Regards,
Bill


-- 
William D. Herrin ................ [email protected]  [email protected]
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to