[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RRG] Topology that follows addressing



In einer eMail vom 22.11.2007 10:33:37 Westeuropäische Normalzeit schreibt iljitsch@muada.com:
I understand this to be geographic routing, which goes even further 
than the geographic aggregation that is pretty much rejected out of 
hand whenever it comes up within the IETF (which is every couple of 
years or so).
But this one is different, and that's why ( my guess) you are mistaken ,next.


Unless I'm mistaken, the idea is that geographic proximity equals 
topological proximity. I'm afraid that can't work in a wired network. 
If I'm in Holland I'm closer to Scotland than when I'm in Belgium. But 
if I want to travel to Scotland through the channel tunnel, I need to 
go through Belgium. So I need to get farther geographically to get 
closer topologically.
The link weight for each loose =hierarchical link is equal to the number of hereby required physical hops.
Hence stretch is equal 1, and in case your router in Scotland is just one hop away than
If your router in Scotland is just one hop away than you are "closer to Scotland than to Belgium" though being in Holland. Nira works like in an OSPF-network, however whereby the network is deminished the more the nodes are further away.

>The way to solve this would be to lift the geo/topo alignment 
>requirement as soon as you reach a certain zoom level.
 
As soon as the current router is marked with the same geo-date as is the packet's destination, any other classic method shall apply (OSPF or BGP).
  
But then 
there's the traditional argument against geographic aggregation: how 
do you avoid ISPs having to carry traffic for free? In other words: 
who announces the aggregates to the rest of the world?
 
You mean to the rest of this geo-patch (remember: no worldwide routing churn! ) ! Well just like it is done in OSPF. Hereby, across any two adjacent ISPs, aggregates shall be announced : 
The representative node from ISP A which has a loose=hierarchical link to some representative node from ISP B shall indeed announce not only his own reachability info, but also that one of all its surrounding ISP-A- nodes, which are closer to it than to any other representative node. 
 

>Are you familiar with Tony Hain's draft about geographic addressing?
No, I am not. But believe me, I am absolutely in accordance with link-state-protocols. I am not in favor of exploiting country codes, nor of any other geo-political information. The goal is just to "thin" the too big internet. If I want to travel von Munich to Manhattan 42 street, I do not care about any road blockage as long as I haven't reached New York ! It wouldn't impact by decision to make the best next hop towards the airport for departure.
 
 
Heiner