[CUWiN-Dev] IPv6 -- TAKE II: NEED ANSWER BY TUESDAY NOON...

[David Young]

On Mon, Apr 25, 2005 at 06:34:02PM -0500, Sascha Meinrath wrote:
> David wrote:
> >A /32 is not necessary or sufficient to ensure unique IPs.  The way that 
> >we will ensure unique IPs is by duplicate address detection/avoidance. 
> >In the long term, we cannot avoid DAD/DAA on ad hoc networks, be they 
> >IPv4 or IPv6 networks.
> 
> Right -- but with a /32 we can ensure that non-uniqueness is due to 
> subterfuge rather than multiple devices being assigned the same IPs 
> because the numbers are quasi-unique hashes from MAC addresses, right? 
> The /32 doesn't eliminated the need for DAD/DAA, but should be sufficient 
> to lessen duplicate addresses as much as is possible.

Trust me, a /32 is not the way to solve the problem.

> >One idea we have bandied about is IPv6 multihoming and 
> >provider-independence.  I don't think a /32 helps us here.  One thing 
> >about IPv6 that seems settled is that provider-based addressing is "in," 
> >and geographical addressing is "out."  In the multi6 working group's 
> >conception of v6 multihoming, each host in an N-homed networked will 
> >have N addresses assigned, one from each provider's prefix.  Through 
> >some protocol to-be-defined, a multihomed host will tell its peers when 
> >it changes from one provider's to another's prefix.  It's all pretty 
> >vague, but it doesn't involve provider-independent addresses.
> 
> I don't follow how a /32 doesn't help -- what would be helpful in this 
> case?

Standards for multi-homing and hosts that implement the standards
will help.

> Frannie and I are working on getting the /32 paid for by another 
> institution (the Public Interest Registry -- www.pir.org) -- ideally, they 
> would be a partner in this endeavor.  I'm certain that there are more 
> ideal solutions than a /32 (and am interested in hearing what they'd be); 
> but I'm also unclear if folks are saying that this is something we 
> _shouldn't_ move forward on, or whether it's more of a "we should do this, 
> but we're not done yet" sort of analysis.

I am saying a /32 doesn't solve the problems you mention (multihoming,
address duplication).  I do not think you should ask for numbers to be
assigned by ARIN.  Numbers will come from our ISPs.

Dave

-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933