Re: Ordered delivery of EAP messages

RADIUS could conceivably reorder packets even where there is
no EAP 
retransmission.

This would seem to be most likely in trans-continental
roaming scenarios 
where the RTTs could be quite large, and the RADIUS RTO is
set low (e.g. a 
second or less) without backoff.

Trans-continential roaming test have shown very high failure
rates for EAP 
conversations with lots of roundtrips, so I wouldn't be
surprised if this 
issue was showing up in real networks.

One way to make this less likely would be for the NAS to
detect FRTO via 
Event-Timestamp & Identifier change, and then wait until
sending a new 
Access-Request, to allow the retransmitted Access-Request to
drain from the 
network.


Alper said:

The problem scenario requires EAP-layer retransmission,
correct?
Authentication server does not perform such retransmission.
So, I don't see
equivalence between the two legs of the EAP transport.


____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap

Arhives: http://lists.
frascone.com/pipermail/eap

> RADIUS could conceivably reorder packets even where
there is no EAP
> retransmission.

Assuming that per session there can only be one outstanding
request for an 
EAP conversation since the conversation depends on acks from
its peer to 
continue.

Then if new messages and retries are sent out in any order
on the wire 
they would either be detected by either side as duplicates
or processed 
and handled accordingly regardless of order of receipt.

take care,
Peter
____________________________________________________________
_____
To unsubscribe or modify your subscription options, please
visit:
http:/
/lists.frascone.com/mailman/listinfo/eap

Arhives: http://lists.
frascone.com/pipermail/eap