On Tue, 25 Sep 2001, Sean Donelan wrote: :But percentage of routes is just one way to measure "importance." :It may not be the best way. Other methods include : : 1. Number of stock options owned by Very Important People : 2. CAIDA skitter traces of routers of confluence : 3. Number of OC-192 links in a building : 4. Number of "Tier 1" providers in a building : 5. Government fiat : 6. Wait for the building to fall down and see what happens Is there a geometric method of measuring the 'meshedness' of a given set? If you take all the as-paths from a sampling of peers across the Internet, and show the relative density of where the respective paths converge, you can get a good picture of who's transiting the most routes. Now this doesn't show physical connections, as an AS can represent an area that spans continents, but it shows who is responsible, which in any DRP/BCP is among the first things established. :Assuming there are locations more impotant than others, should :we do anything? Or should we just hope no one else figures out :where they are? Well, the gov, or an industry consortium can find these dense transit areas and require that organizations with (eg.)2 or more peers have some semblance of a DRP/BCP plan that can be audited. The plan doesn't nessecarily have to garuntee connectivity, but establish whether they can be trusted to route packets if the DRP/BCP has to be initialized. After all, we are talking about the Internet, and though many orgs control lots of different parts of the infrastructure, maybe a plan just for layer 3 might be worth persuing. So maybe a large percentage of traffic gets routed through 60 hudson, the providers that are located there would each have to have diverse enough infrastructure/routing policies to contend with the unavailability of their equipment at that facility, to qualify as an Infrastructure Carrier. In short, I think that a plan like this should start on ground we all know and have the power to negotiate on, which is layer 3. -- batz Reluctant Ninja Defective Technologies