[ppml] [narten at us.ibm.com: PI addressing in IPv6 advances in ARIN]

[David Conrad]

Chris,

On Apr 16, 2006, at 10:10 PM, Christopher Morrow wrote:
> There is no way to get back the /48's assigned under PI
> space after a multihoming solution that makes sense arrives, and there
> is, honestly, no driver to continue working on a multihoming solution
> now that there is PI space.

I tend to view multi-homing as a subset of a more general problem,  
that of separation of end point identifier and routing locator.   
Renumbering is another aspect of this problem. Mobility too. PI is  
just an attempt to ignore the problem and hope it goes away.

I suspect the PI approach is a short-term solution to multi-homing/ 
renumbering: when ISPs start feeling pain, they'll take whatever  
steps they feel necessary to protect their infrastructure, just like  
they did with IPv4.  History repeats itself.  It'll be a bit more  
difficult with IPv6 than it was with IPv4 as prefix-length filters  
won't work as well (given everybody gets a /48), but I'm sure smart  
folks at ISPs will come up with other (unfortunately, likely equally  
annoying from the perspective of end users) solutions.

> it seems that backing up, restarting the 'new protocol' process is
> likely to end up with another 10 years wasted, so it's very hard to
> see a reasonable path forward at this time.

As far as I can tell, there are two solutions to the fact that IPv6  
didn't separate end point identifier from routing locator:

1) the shim6 approach: create a layer 2.5 solution that separates end  
point identifier and routing locator and re-implement IPv6 on all  
hosts to make use of that layer;

2) the middlebox approach: separate the end point identifier from the  
routing locator at the core/edge boundary, leaving the IPv6 stacks at  
both the routers in the core and the hosts at the edge untouched, but  
inserting a mapping from end point identifier to routing locator on  
source edge to core and an unmapping the routing locator back to the  
end point identifier from core to destination edge.

(I'm not smart enough to figure out how the geotopo approach would  
work without causing a return to the equivalent of the pre-1994  
international telephony settlements regime, but that's my own lack of  
imagination I suspect).

I personally think the middlebox approach is the easiest to deploy/ 
least disruptive to end users/most familiar to ISPs technique to  
implement an end point identifier/routing locator split, but I'm  
cynical enough to be skeptical either approach will be taken...

Rgds,
-drc
--------
My opinions are my own and do not necessarily represent the
opinions of any organization I may be a part of.  So there.