In einer eMail vom 11.12.2008 23:11:52 Westeuropäische Normalzeit schreibt [email protected]:
On Thu, Dec 11, 2008 at 3:42 PM, <[email protected]> wrote: > Some comment and proposals to improve the strategy C text which currently > is: This references http://bill.herrin.us/network/rrgarchitectures.html The current text is: > Suppress distant routes by aggregating them into sets expected to be > available in a given direction. Because LOC reachability info is not > flooded, the routing tables each router must deal with are relatively small. You should say, because user reachability information is not flooded,.. > --------------------------proposed text: > Extend BGP such that routers can acquire the view of a well-sparsed internet > topology > with strict links in the near surrounding and - in general - with loose and > looser links the > more remote they are (strict links to very remote nodes may still be part of > this topology). Determine the next best hop based on that viewed > destination node which is either the true destination node or a node which > is closest to the true but not yet visible destination node. Arrange the > results such that a best next hop can be retrieved either by 1 or by 3 > table lookups. As I mentioned in our email exchange, I don't understand what this means. What is a "well-sparsed" Internet? A topology which doesn't show the details of the remote network parts. Similar to routing from NY, Broadway to Sausolito, Main Street without seeing the Main-Street prior entering Sausilito itself. What is a loose or strict link? This is IETF terminology. How do you determine that a node is "closest to" a node that isn't visible? By the distance (in miles or km) which is adequate in location base routing - IMO. What is the significance of 1 or 3 table lookups? Much faster forwarding. Eliminating the need for caching. > Of course, incremental deployment has to be supported.The goal is to shrink > the routing table continuously so that it becomes empty as soon as all DFZ > routers will comply. This is a goal, not a strategy. I skipped the previous explanatory sentence because I thought it is too hard to understand :-( I have one further comment with respect to the mentioned economic model of IP. It is just the opposite: Strategy C avoids everything which looks like administrative hierarchy. Rather ask ILNP or HAIR or Joel's presentation. There is no single hint for any political/administrative influence in all of what I presented. Another point: I have always argued against the generalization that hierarchical routing is doomed to cause stretch=3 because it is not true wrt TARA. However I can well understand that this reproach is appropriate to ALL the currently arising hierarchical ideas. Heiner
_______________________________________________ rrg mailing list [email protected] https://www.irtf.org/mailman/listinfo/rrg
