RE: Location

["Bound, Jim" <Jim.Bound@hp.com>]

>>Location: 

>> Do we believe that providers should know location of all nodes on the
Internet >>without peering? 


>That implies that you can route based on static location and implies
that your 
>locators are independent of the network topology.  Unfortunately, that
is 
>exactly the reverse of what scales.  It implies that you must flood 
>topology information for the entire net.

I agree.  I do not believe this is a solution but put it here to kill
it.


>>Do we believe that providers should know location of all nodes within
a 
>> specific topology, without peering? 

>Same answer.  You either distribute full topology information or prefix

>information.

I think answer is different.  If the topology sectors are say under
ordinal 100?
Why is that unreasonable.  We do not need all topology flood for each
topology sector or AS domain.  It is a simple fast path to 100 or less
locations?  I believe we can begin the locations less than 20 initially
for IPv6?


>>Do we believe that providers should know location of all nodes within
a 
>>specific topology within specific rendezvous points within the
aforementioned >>specific topology, without peering? 

>I can't parse.

If location became greater than 100 then within each topology some
locations become redezvous point (RZP).  The RZPs are not propogated
above any topology sector. It is hierarchical lower route only to that
location?  

>>Do we believe that providers should only know peerings for providers
that 
>> exist near specific locations? 

>This either.

Provider = Topology Sectors.  Providers would not know RZPs in other
Sectors?

Thanks
/jim

Tony