Getting a "portable" /19 or /20

Wed Apr 11 03:02:52 UTC 2001

On Tue, 10 Apr 2001, Sean M. Doran wrote:

> Greg Maxwell <gmaxwell at martin.fl.us> writes:
> 
> | > 	Aggregation buys time, that's it.  Aggregation does not make the
> | > current routing methods any more scalable.
> |
> | In IPv4 yes, because you can't have perfect aggregation, too much network
> | multihoming and old prefixes and it's to painful to change address blocks.
> |
> | In IPv6, if implimented right aggregation provides for virtually limitless
> | scalability for unicast traffic.
> 
> Perfectly aggregated networks are star-shaped.  

No. They are tree shaped.

> Any more complicated topology cannot be perfectly aggregated.

>From what perspective?

A bunch of seperate funky meshes which each only advertise a single prefix
to me looks perfectly aggregated. 

How do you isolate the mess? You use global keys which convey ASN specific
path knoweldge. Like the post office, you need not know what a part of the
address means unless you are a participant in routing that part. So you
make every ASN represent their network as a perfect tree to the outside.

So how do you do things like multihoming then, which is what (other then
cruft) breaks the tree in IPv4, you stop doing it in the network layer. It
doesn't belong there, you place it in the transport and make it
end-to-end.

Non-top-level providers pass on prefixes for each of their upstreams,
their users inherit those and the ones from their other providers. This is
not likely possible (or at least sufficently easy) in IPv4, but is totally
reasonable in IPv6.

> In real networks, aggregation at best follows a "reasonable"
> trade-off between optimizing and stabilizing route selection.
> Not everyone will agree on what is a "reasonable" balance.  

That can be changed by removing path selection from the routing process.
(I differentiate 'path selection' here from 'route selection', when you
decided FedEX vs USPS you are performing path selection, when FedEX places
your package on plane 123 it has performed 'route selection')

> Result: some people unhappy about suboptimal routing ("my packets
> to my neighbour across the street go through another country") and
> some people unhappy about too-great dynamicism ("damn, time to upgrade
> to a faster processor, more memory, faster memory, etc etc etc").

Go through another country if it's two hops across an OC192. :) 

I don't think that every network should be perfectly aggregated, I think
that every network should advertised as though it were perfectly
aggregated. This perfection from the global perspective will
free up so much routing table space that everyone except the largest
providers will be able to be totally unaggregated internally allowing for
optimum routing. 

Hosts will then either be multihomed and be able to choose which of their
links is closest to the target network or cope with the climb up the tree
to reach their destination.

> This is a result of the CIDR addressing architecture and is 
> INDEPENDENT OF THE NUMBER OF BITS IN AN ADDRESS.

I disagree, if insted we didn't aggregate into /19s or /20s but into /22s
and /23 there would be less unoptimum routing and larger (less scalable)
route tables.