>         And application break when you use non-globally-routed GA's.
> 
>         People want permanent private address space be that multiply
>         allocated (SL/1918) or non-globally-routed global address
>         space (NGRGAS).  Passing addresses in either is not
>         guaranteed to work with every node.  

no.  but if you use globals for all addresses then you can get a
reasonable separation of function between apps/hosts and the network -
if you can't communicate between two nodes it's either because part of
the network is down or because the network doesn't let you.  either
way, it's not something that the application can be expected to solve.
so with globals the application no longer has to be responsible for 
trying to work around deficiencies in the addressing scheme.   

>         NGRGAS removes one
>         address maps to multiple nodes 

yes.

> but it does not help identify
>         if two nodes share a common site though you can identify
>         if a node shares your site.

the node shouldn't need to do this.  the app sends to an address.  
if it gets there, it gets there.  if not, it's not the app's problem.
the app can try to fail-over to other addresses if they exist, but
the first address should work with high probability.  in other words,
we need to abandon this idea that a host can have an arbitrary number
of addresses with varying scopes and lifetimes and connectivity, 
and it's up to hosts/apps to sort out which one is best to use.
instead hosts should have a small number of addresses (most of the time
just 1, 2 during renumbering periods) each of which is valid from 
anywhere, and the network should decide how best to get the traffic 
there.  (no I'm not suggesting to deprecate LLs but normal apps should 
use them only as a last resort or when explicitly configured to do so.) 
 
>         NGRGAS also requires a registry to allocate the address space.

if we tried I'm sure we could come up with a way to allocate non-routable
globals without a registry.   I'm convinced that this is vital for the
long-term success of IPv6.

Keith
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to