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

Re: [RRG] Idea for shooting down



In einer eMail vom 21.11.2007 00:43:31 Westeuropäische Normalzeit schreibt tli@cisco.com:

So would it be fair to characterize this as a locator/ID split, where both locator and ID are present in the packet header, and locators are geographically assigned?  
You may or may not see it as a different type of locator/ID split. It is first of all a scalable way for routing.
Wrt. ID: In the years 2013/2014 we might be glad to continue with IPv4 because it would provide enough addresses which are unique within one geo-patch. 
Routing within the location is doing by normal hierarchical routing.
Let's say by traditional reachability-info-based determination of the egress node.

As always with geographic abstractions, how do you deal with the case of two sites within the same geographic granularity that have no interconnectivity except outside of their locale?
Partitioning is always a troublemaker :-). One way to deal with is to cheat: If I remember correctly, then not all islands of the Hawaii island group are in the same time zone, but for practical reason they have all
Honolulu time.
But of course, this problem should be dealt with by the protocol: We may do something which allows us to detect e.g. that Alaska (which is truly the case) has a partitioned road system, e.g. by recognizing that all its representative nodes in the hierarchically next upper level (country map) are more than those that the current Anchorage node can discern on its state map.Yes, it may involve a crankback detour.
This problem would also show up in case of  partitioning due to some link failures. I am much in favor that any viewed topology is the result of a routing protocol and not of any configurational data.
Configurational, IANA-based info should only guide the way how the hierarchy evolves, i.e. which lower geo-patches should be combined to the next upper geo-patch.
Political borders must not have any influence ! However IANA may optimize the  geo-patches,e.g. as to get larger geo-patches on the oceans than on soil.
 
But there may be further ideas how to deal with partitioned geo-patches (have a number for each part ?!)
Note: Because the hierarchical network view is the result of a routing protocol the partion may also hit any larger geo-patch at any upper level (partitioning of a "country" needs to be handled, likewise.) 
 
Heiner