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

Re: [RRG] Convergence



In einer eMail vom 04.12.2007 01:37:18 Westeuropäische Normalzeit schreibt tli@cisco.com:
I should remind folks that our 'problem' is a matter of research, not 
engineering. 

Tony
This is why my emails and proposal have been addressed to RRG/IRTF.
 
Imho, LISP should go the usual way: have IETF-BOFs, initiate a LISP WG inside IETF.
IRTF stands for research however.
Hence RRG should not ask in the first place about the backward compatability, instead AFTER having determined, independendly, what is the best solution.
 
So far I did dispose all about the NIRA-concept which :
- eliminates the scalability problem: it reduces the routing table size immensely. It stops the routing churn:
In case any representative node went down the node the nearest among all nodes being BoundTo the failing node would take care if there were any "taking care on behalf of"-problems. A new update to the outside is only appropriate, if the node who is in charge of some other has a distance > xyz to that failing node.
 
-  enables faster packet forwarding by factor 20 (according to Tony).
-  only needs the avaibility of longitude/latitude of the packet destination in one way or the other
  (parameter, outer header,...) but no CONS and no cumbersome config data. 
 
- by knowing the (sparsed) internet topology you can do QoS/policy routing much better than without.
  The goal should be not only as well as inside an OSPF-network but even better (OSPF only knows to deal 
   with Dijkstra) 
 
- it enables to deal with the multihoming problem in a new way, as well as with traffic balancing issues. 
 
I have been asked for providing a draft about this concept by several people.Where are their comments?
 
Heiner