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

Re: [RRG] Topology that follows addressing



Philip, thanks for your email.
See my comments inside whether your understand does or does not comply with what I was writing.
Heiner
 
In einer eMail vom 09.11.2007 16:56:27 Westeuropäische Normalzeit schreibt philip.eardley@bt.com:

Heiner

 

Thanks. re your first para - personally I find it very useful to understand what a protocol is trying to do at a high level and why - before reading all the detail.

 

Let me see if I understand your approach:

- A node’s address is simply its geographic location [based on latitude/longitude].

No. The geographical location is only an attribute. Within the same geographical area there are, in general, multiple routers which show up at some more upper map. You might object, well, a single particularly elected router shall represent let's say a fifth of all router's from California. What if that one goes down?
Well, all other routers which are closer to this one than to any other representative router from California, would continue to "stick to it".
Sometimes ugly situations like such one are helpful to demonstrate the bottom line:
If I am sending from Munich, Germany, a postal letter to someone in Sausolito, who however has moved to a different location, it is perfectly fine - in my view - that this letter is first of all sent to some postal office in Sausolito rather. For me, it is acceptable that such a letter has to be sent back all the long way because I believe, it makes no sense, that if someone relocates, all postal offices in the world should be informed. Also, in that case we wouldn't have got the lovely Elvis song RETURN TO SENDER :-)
 

 

- The topology is hierarchical & follows the addressing.

The topology is hierarchical and the way the hierarchy is "formed up" is according to geograph. attributes.

Geographic-based routing is an interesting idea and there has been some study of it in the ad hoc community I believe. Couple of questions:

- Would any policy control be possible?

 For sure, because as soon as some concept is established people are smart enough to enhance it :-) 
Also: Imagine, that your OSPF routers would also have these geo-attributes! You may try to get closest to the destination inside your OSPF network, e.g.

eg to route through a particular higher level network that you have a commercial deal with. How about TE & multihoming?

How about multiple routes with even different   hierarchical links towards the egress?!
 
All these things would seem to me to make it more problematic to enforce ‘topology must follow geography’. in the context of a global network, would it mean there was only one network?
 
 

- how would failures be dealt with? Can you discover & swap to and alternative path? (this would then mean that topology didn’t follow geography)

Maybe it is better to put this law aside again, there is no extra value in following this or that. The proposed concept was not developed based on this law.

- could you handle mobile nodes and networks?

We can cross that bridge when we have arrived there.
 
Heiner