The IP Over NBMA (ION) Archive

[gardo@VNET.IBM.COM]

Matthew,

> > > o Should holdtime for a NAK be less than that of a normal reply ?
> > The "holdtime" in an NHRP packet is zero when it is a NAK.  Maybe you
> > were not referring to "Holding Time" in the packet.
>     Well spotted - the main aim of this was to consider the rate at
> which we send Requests (limited by whatever means). Given the current
> spec, what rate do we send Requests for the unreachable destination ?
> More importantly, how do we want the Requester to behave ?
>
>     At present, some implementations might send Requests every 60
> seconds, while others would send them more frequently. We need to send
> them often enough (or have another mechanism) to ensure that we detect
> the recovery of a route "fast enough". Too many Requests would be bad
> news though.
Obviously, the "routed path" recovery time is not influenced by NHRP.
However, I think the holding time in the Resolution Reply *could* be used
to minimize NHRP Resolution Replies with NAKs.  The NHS *could* simply
send a purge for destinations that changed, including destinations
that were NAK'ed if they change before the Holding Time that the NHS
put in the Resolution Reply.  Zero could still be used to mean that the
NHS will not send a purge.  This would be a text-only change, no
packet format changes or functional changes....

Currently, an "unreachable destination" causes an NHRP Error Indication
packet to be sent in response to a Resolution Request when the
destination is unreachable;  the requester does not receive a
Resolution Reply with a NAK in this case, at least that's my
understanding....  I'm sure someone will correct me if I'm wrong!  :-)

Have a nice day!
-- Russell