...

I'm not saying CIDR isn't going to be highly beneficial, it will be. I'm just saying that Tony's figures are optimistic.

This brings up one of the items I wanted to talk about in ALE next week: the figures that I've been generating for the growth charts to estimate how much CIDR helps are also based off Merit's nets.unl files, but look at only the first AS number to decide where the aggregation occurs.. Would that be considered overly pessimistic? -- Frank

Frank, I'm sorry but I haven't been keeping up with ALE activities. There are almost 28,000 nets configured and less than 19,000 announced. This means that 9,000 networks are going configured portions of CIDR blocks that are not yet in use. You probably want take that into consideration when reporting percentages gained by CIDR. Failing to do so would overstate the benefits of CIDR. Using the border AS number is just plain inaccurate. A better (but still not perfect) means of making this estimate would be to take a routing dump and compare the entire AS path. For example, if Tony (or someone in Europe) has two routes for AS 86 and one goes through AS 287 before hitting AS 690 and one goes straight to AS 690, he knows that AS 690 could proxy aggregate in the future (since AS 287 will go away shortly). If the ANS fake AS kludge (technical details on request) where not there, you would just see AS 86, but proxy aggregation would still be required to not break Suranet's load balancing acroos there Washington and Atlanta attachments. There may be other cases of two providers that touch in two places that you can't see looking at a routing dump. ALE is entertaining, but I don't think you will be able to get very accurate estimates of CIDR benefits. I'll get you a gated dump from the ANS backbone, if you need to entertain yourself. :-) Seriously - Let's take this discussion of how to make the estimates more accurate off line or move it to the ALE list (there is one right?). I will get you a gated dump if you want to do an analysis based on it, or you can get a routing dump from someone in Europe (or both). My view of CIDR is that we absolutely need to get moving on replacing more specific routes with aggregates as soon as we can be assured it won't break things. Then we can see how much we are able to initially reduce routing table size and project from there. Curtis BTW - aggregation doesn't occur by magic. If the very tiny number of aggregates registered at Merit through NACRs is any indication of how much effort providers are willing to put into figuring out what they can aggregate, CIDR isn't going to buy a whole lot. Please start registering aggregates you intend to announce so we have a good basis for estimating a lower bound on what CIDR _will_ buy us (as opposed to "could theoretically"). There is no harm in having ANS configure the aggregate and the more specific routes and hear only the more specific routes until the bugs get worked out. (Hey - it's only memory on the routers - we'll just plug in 64MB were we need it:).