The MPLS WG Archive

["James R. Leu" <jleu@mindspring.com>]

My understanding is that the list of addresses messages given to me by a
peer will allow me to answer the question if this peer "is the nexthop" for
a particular route.  This allow me to know whether or not I should install
a mapping the peer has given me.  What if ... the address that I match
(which was given to me by the peer) is from a differnt labelspace then the
label he has allocated to me?  Maybe a picture will help.


   ---------                 ---------
  |         |A11---------A21|         |
  |   R1    |               |    R2   |------ FEC
  |         |A12---------A22|         |
   ---------                 ---------

-A21 means R2's 1st address.
-each interface has a labelspace that matches the address number
 (thus R2 has 2 labelspaces it allocates from 21,22)
-R1 and R2 have two LDP sessions
-R1 session with R2:21 has A22,A21 for an address list
-R1 session with R2:22 has A21,A22 for an address list

-R1 has a routing table of:
 Prefix     outif      nexthop
 FEC        A11        A21

R2:22 allocates a label(L1) in labelspace 22 and send the mapping to R1.
R1 gets the mapping and searches the address list for R2:22 and finds that
the next hop of A21 is in R2:22 address list, so it installs the label out
A12 for FEC.

-R1 now has a FTN that looks like:
 Prefix     outif      nexthop   label
 FEC        A12        A22       L1

I can see avoiding this problem if the check for "is the nexthop" also takes
into consideration the outgoing if.  How would that work in targeted mode?

I think this can be solved for both modes by limiting the address messages 
R2:22 gave to R1. Why not only give out the addresses of the interfaces that
match the labelspace you are advertising?

Jim
-- 
James R. Leu