[ppml] 2005-1 status

[Scott Leibrand]

Owen,

I speculated that renumbering difficulty is "a function of the number of
IPv6 subnets in use and the number of statically configured IPv6 hosts".
I did not (and do not) assume that renumbering difficulty is a direct
function of the number of devices.  I recognize that (stateful|stateless)
autoconfigured devices are very easy to renumber, whereas statically
configured devices, which may have various references pointing to the
static IP that must be updated, are much more difficult.

I'm glad I'm not the only one having difficulty quantifying renumbering
difficulty.  On the other hand, if someone *has* come up a good way to
quantify renumbering difficulty, please do share it.  In the absence of a
better method, I can accept that efficient utilization of ~500 IPv6
addresses would be a reasonable threshold for getting PI space if that's
the threshold favored by others.  I would also accept a composite metric
based on a lower number (say ~200) of statically configured addresses, or
a similarly smaller number (maybe ~100) of unique subnets (discrete
physical broadcast domains).  Or perhaps meeting any one of those three
thresholds would be adequate to get PI space.

What threshold(s) would everyone else consider reasonable?

-Scott

On 01/24/06 at 2:33pm -0800, Owen DeLong <owen at delong.com> wrote:

> Scott,
> 	You continue to accept the fallacious assumption that
> renumbering difficulty can be measured by prefix length or number
> of devices.  This simply isn't true.
>
> 	Here are some of the things that can make renumbering difficult:
>
> 	+	Local addresses embedded in remote device configurations
> 		*	ACLs at customer/supplier/contractor sites
> 		*	VPN configurations
> 		*	NS Glue records
>
> 	+	Number of statically configured devices
> 		*	Routers
> 		*	Servers
> 		*	Others
>
> 	+	Number of users affected by "flag day"
>
> 	+	Local addresses specified in zones not controlled by
> 		local administrator
>
> 		*	Web Hosting
> 		*	Certain Application Service Provider situations
> 		*	eMail hosting
> 		*	Other forms of server outsourcing
>
> 	Notice that most of them are affected by the number of remote sites
> involved far more than they are affected by the number of local
> addresses
> involved.
>
> 	I have not, as yet, come up with a good method for quantifying
> renumbering difficulty.  I can accept efficient utilization of 508
> addresses as a reasonable compromise yardstick if you can (current IPv4
> policy).  If you can't accept that, please elaborate on what you think
> would be acceptable.
>
> Owen
>
>