RE: Reducing Peerings for MH Routing within a site via end systems

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

Tony,

Thanks.  No response will try to not do the pronoun thing.

/jim

 


> -----Original Message-----
> From: Tony Li [mailto:Tony.Li@procket.com] 
> Sent: Monday, March 17, 2003 11:20 PM
> To: Bound, Jim; multi6@ops.ietf.org
> Subject: RE: Reducing Peerings for MH Routing within a site 
> via end systems
> 
> 
> 
> 
> Jim,
> 
> |    >If a host has multiple locators and a single identifier 
> |    and other hosts
> |    are able >to change the locator that they are using at 
> |    will, then it
> |    implies that special
> |    
> |    Clarification.  Above you have "conjunction" ...and 
> other hosts?  I
> |    assume you mean any host can change their locator?
> 
> 
> I think that a host should be able to change the set of 
> locators that make it reachable at will.  The rate that the 
> system can disseminate these changes is still TBD, but there 
> obviously needs to be an upper bound (no pun intended ;-).
> 
> Correspondent hosts are free to try different locators at any 
> time.  Typically they will use transport layer hints to 
> trigger things.
> 
> 
> |    >Effectively, the site has PA space from each of its 
> |    providers and that
> |    PA space >is nicely aggregated in global routing.
> |    
> |    So the locators for this site is known by all of the sites 
> |    providers,
> |    correct?
> 
> 
> Yes.  Even more strongly, the locators are allocated by the providers.
> 
> 
> |    IPsec between those two peers without any translation of 
> |    the packets
> |    headers.
> 
> 
> Ipsec, in particular, needs to use the identifier in crafting 
> the security association, NOT the locators.  [This would also 
> allow Ipsec to work through NAT.  ;-) ]
> 
> 
> |    Do we agree on that and this is a core principle for any Internet
> |    engineer one way or the other?
> 
> 
> Please restate with dereferenced pronouns.
> 
>   
> |    >One could reasonably argue that you take the layer of 
> |    separation down 
> |    >recursively.
> |    
> |    Within the IGP from the Provider space?  If that, then I agree.
> 
> 
> Within the site IGP, but using the locators provided in PA space.
> 
>     
> |    >Note that this alleviates the most painful reason to 
> |    avoid PA space:
> |    
> |    But above you state "nicely aggregates provider space"?  Can you
> |    rectify?
> 
> 
> These aren't incompatible.  Your locators are aggregateable 
> and you don't need to renumber.  Should make everyone happy...  ;-)
> 
>      
> |    >renumbering is now a non-issue.  You assign a new 
> locator to your
> |    entire site, 
> |    >advertise it to the world and you're good to go.  
> Individual host
> |    identifiers 
> |    >need not change.  No host renumbering.
> |    
> |    But still implies locators can change?  For example one move to
> |    completely new providers?
> 
> 
> Of course.  You add new provider, configure edge router with 
> new locator prefix, update the locator database and you're done.
> 
>     
> |    Your location is not worldwide provider space but only 
> for relative
> |    provider space for a site?
> |    Is that correct?
> 
> 
> Locators are part of what we would now think of as PA space.  
> That is, they are topologically significant information that 
> vary per-provider.
> 
> Tony
>     
>