Policy Statement on Address Space Allocations
Noel Chiappa
jnc at ginger.lcs.mit.edu
Sat Jan 27 17:06:15 UTC 1996
<I'm mostly sitting this one out, guys and gals... Been there, done that,
have *all* the T-shirts, way too many times... I know I should really just
ignore Tim, sending him replies just keeps him going, but this one was too
food to pass up. For now I'll be totally civil, and let Sean handle the
ridicule and abuse approach... :->
From: Tim Bass (from cais.com) <cais at dune.silkroad.com>
There are at least five hierarchical routing architectures that are
not CIDR based
To start with, CIDR is not a "routing architecture", it's an addressing
scheme; routing architectures are pretty much orthagonal. I.e. you could run a
hierarchical link-state system sort of like OSPF in the Internet with the
exact same addressing as we have today (of course, you'd have to add more
levels of areas to the protocol, etc).
So, assuming you meant "five hierachical addressing schemes", could you
enumerate them for the rest of us? I can think of only one addressing scheme
which is really significantly different from CIDR (Landmark doesn't count,
it's basically CIDR with diffuse boundaries), and that's Geographic.
What is hard to understand is why men and women with intelligent brains
believe that there is only 'one way, one religion' to do hierarchical
routing.
You need to distinguish between routing, and the addresing structure it uses.
There are numerous ways to do the routing (i.e. the calculation of paths
and/or routing table entries, such as destination-vector, link-state, etc),
but there seems to be only one way to make the addressing work for very large
networks, and that is to use the approach in CIDR.
I am quite sure that the most vocal advocates of CIDR have read very little
historical documents outside of the IETF RFCs.
Oh, really? I guess I don't count as a very vocal advocate of CIDR... and
I guess the same goes for Tony and Yakov. Hi guys! :-)
I highly recommend some time spent in review of the principals of
hierarchical routing; graph theory; queuing systems; dynamic programming
and others as well as the marginalia of journal articles on the subject.
Queueing systems aren't much use for this aspect of routing.
Noel
More information about the NANOG
mailing list