One of the advantages of SKIP claimed is that gateways can be configured
in parallel to perform instant-failover. By this, I envision that if you
have 3 networks, A B and C, you can enter skiphost records to create
multiple tunnels for each destination network.
Thus on host A you would have net B -> tunnel AB and net B -> tunnel
AC. The primary record on host C (net B -> tunnel CB) would then forward
the packets on.
Is this a correct understanding? If both tunnels are available how does
SKIP know to choose the most direct route?
Also, if each of n hosts with a choice of n-1 paths using redundancy
r (r<=n-2) additional paths will have (r+1)*(n-1) skiphost entries. 10
nodes of a VPN using all possibilities would require 81 skiphost entries
on each skiphost and the VPN system would require 810 entries if all 10
skiphosts were configured the same.
Does this sound logical? Looks like a good case for careful planning of
the redundant entries to limit them to one or two per node.
Jim Flowers <[EMAIL PROTECTED]>
#4 ISP on C|NET, #1 in Ohio
