Resolving geo discussions

[Iljitsch van Beijnum <iljitsch@muada.com>]

Last week we had another round of "this can't work" / "yes it can" 
discussions about geographic aggregation. While doing this from time to 
time is a lot of fun, I'm sure we can all find more productive things 
to do with our time.

So how do we resolve this in a way that is satisfactory to everyone?

The way I see it, many of the "con" arguments may turn out to be quite 
valid. However, there are also many ways in which geo aggregation could 
be successful. Maybe it will solve the multihoming scaling problem for 
10, 5 or just 2 years. Or it just helps regional networks to optimize 
their routing. I don't think there is any way to determine this 
beforehand.

But it should be possible to adequately predict the costs and risks in 
advance. The costs for the internet community as a whole mostly consist 
of the RIRs having to implement tools to make it possible to assign 
geographically aggregatable PI addresses to multihomers.

The only risk that I can see is that the GAPI addresses will in essence 
turn out to be nothing more than regular PI addresses. This means an 
explosion of the global routing table, and subsequently at some point 
networks will start to filter out GAPI more specifics.

Is there a way we can arrive on some consensus as to whether the 
(potential) downsides are worth the potential gain?

We could experiment with this for a set amount of time in 6bone space, 
for instance.