I'm sure that we will all eventually do CIDR/BGP4 and maybe even slow the growth of routing tables to a managable level. By this time we will all have 500MB routers..... Now with talk of global renumbering and a Utopian world of CIDR, what do we lose? I was never optimistic of how CIDR would work in a real operational redundant topology like the current internet. In this scenario, it is a big win to be able to have fine grain control of routing down to the organizational level. What does this mean? Well you can develop a routing strategies which take into account many diverse exit points from your AS. You can develop routing strategies based on high end users, low-end users, route the east coast here, the west coast there. You name it. You can't do this by routing large blocks of networks around. Pure CIDR'ists will say that you should divy up your blocks and assign addresses accordingly. Well... how many people who run networks have stayed with the exact same routing strategy with zero changes for more than 6 months? New connections and connectivity happen. The internet is an ever changing place. My CIDR allocation today will not be good for my routing strategy 2 months from now. As Andrew said, we are not fortune tellers... So, In a totally static topology, CIDR "the movie" and global renumbering will work. I wouldn't bet on static topology... Mark
Mark: Just catching up on mail, so I apologize if this is out of sequence.
What does this mean? Well you can develop a routing strategies which take into account many diverse exit points from your AS. You can develop routing strategies based on high end users, low-end users, route the east coast here, the west coast there. You name it.
It sounds like long-range you want a IPng protocol that allows flexible auto-assignment of network numbers? Perhaps someone should mention this to the IPng folks? I know of autoconfiguration work in some of the protocols for IPng. ;-) Sue Hares
participants (2)
-
fedor@msf.psi.net
-
Susan Hares