eap Mailing List: Re: Ordered delivery of EAP messages

[Peter Deacon (peterd]

On Tue, 6 Mar 2007, Bernard Aboba wrote:

> > 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.

> ACK/NAK protocols can have more than one packet in flight, if you take false
> retransmissions into account.

> > 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.

> Both EAP and RADIUS do not detect duplicates arriving out of order; they
> depend on the underlying lower layer to provide in-order delivery so as to
> enable duplicate detection.

A RADIUS server can detect a packet it has already responded to and 
respond if it wishes with a stored response.

A false retranmission by a NAS is detectable by the RADIUS server.

The difference between request and packet is significant in this context.

What I'm saying is in any given RADIUS+EAP exchange there is head-of-line 
condition that prevents two or more unacknowledge requests or responses 
that are *not detectable* by the RADIUS server.  At any given instant one 
or more duplicates may already be known to the RADIUS server and buzzing 
around the wires however only exactly zero or one new requests can be 
outstanding (not known to the RADIUS server)

take care,
Peter