Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email firstname.lastname@example.org
From: Alex Zinin (azininNEXSI.COM)
Date: Thu Sep 13 2001 - 14:42:03 CDT
Answers below, pls.
> 1)Why we must flush the MaxSequenceNumber LSA?
When a router receives an LSA and it has a version
of the same LSA in its LSDB, it needs to decide which
version is newer. If LSA SeqNums are different, the
decision is based on SeqNum comparison and if
(lsa1->seq_num > lsa2->seq_num),
then lsa1 is considered more recent than lsa2. When
an LSA's SeqNum reaches it's max value, the next version
of the LSA would wrap the value back to InitialSequenceNumber
(because of the linear SeqNum space). If we do not make sure
that the MaxSeqNum version of the LSA is purged from
the domain, we will have a problem with the comparison
above and the new LSA will not be installed by the routers.
> 2)In Page 130, in section 18.104.22.168 Option 2, for IP
> numbered PTP link, If I can't get the peer address,
> which subnet will be assigned to the PTP link, my own
> or peer?
You always know your nbr's IP address from the neighbor
data structure, where the IP address is set from the
src IP addr in the packets received from the peer.
> 3)why RFC don't discribe the Receiving the LSU packet
> and Send out the LSU packet, and now I am comfused
> about the receiving LSU?
It does describe it, actually. See Section 13, pls.
> When I received a LSU, I need
> to send out a LSA packet?
Again, see Section 13 pls. Whether you need to send
an LSU out or not will depend on the results of the flooding
> 4)In Page 145, at the point (8), when received a LSU,
> the database copy LSA is more recent, I will send a
> LSU to neighbor, and that will make neighbor router
> NBR state machine to generate BadLSReq event,right?
Nope, the nbr will follow the bullets in (5).
> 5)In the Page 147, when installing the Router-LSAs and
> network-LSAs to the database, the entire routing table
> must be recalculated,the reason is that the AS
> boundary routers may belong to multiple areas.I can't
> understand this sentence,even the router belongs to
> multiple area, it can only run the SPF in one area and
> then generate the type 3 and 4 LSA to other areas, I
> think it is ok.
I'd say this will depend on the actual implementation,
see some discussion on this topic earlier on this list.
The spec defines the safe way of doing this.