On Wed, Oct 10, 2007 at 04:16:16PM -0700, Bernard Aboba wrote:
> >I'm wondering if it can work over wired networks where a
> >non-forwardable multicast address is used as the destination MAC
> >address of EAPoL frames.
> 
> IEEE 802.11i supports 802.1X pre-authentication, in which another Ethertype 
> is used, and 802.1X frames are sent to a unicast destination, forwarded by 
> as many switches as necessary.

This 802.1X usage is specific to IEEE 802.11i pre-authentication where
802.1X exchanged over the wired port of the target AP is used for
controlling wireless ports of the AP.  I don't think we can make it
applicable to DSL simply because of different usage.  I believe there
would be lots of corner cases to tackle with if we define an extention
to 802.1X that can generally work across multiple LAN segments.

> 
> Scenarios involving forwarding of multicast frames are typically limited to 
> situations in which the switch terminating IEEE 802.1X is one hop away, and 
> the forwarding switch acts as a TPMR for 802.1X traffic.  For example, a 
> wired VOIP phone might have a switch port, but does not act as a RADIUS 
> client, so it forwards 802.1X traffic to a switch at the wall-port.

I think that two port MAC relay is not applicable to DSL especially in
bridge mode where multiple Supplicants may exist.

> 
> >How two Supplicants attached to such a
> >switch can run 802.1X where one Supplicant may receive EAPoL frames
> >intended to be received by the other one?
> 
> Typically the 802.1X forwarder will not send the 802.1X frames to all 
> ports, just to the switch one hop deeper in the network.  That way other 
> supplicants should not get confused.
> 

I believe downlink Supplicants directly attached to a switch with more
than two ports still get confused, while the direct uplink
Authenticator would not get confused.

Yoshihiro Ohba



_______________________________________________
Int-area mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/int-area

Reply via email to