[ppml] 2005-1:Business Need for PI Assignments

[David Conrad]

Lea,

On Apr 20, 2005, at 12:06 PM, Lea Roberts wrote:
> David, Randy (et al) -0-

This demonstrates my interest in multihoming... :-)

> I too would like to continue to live the original fantasy of IPv6 where
> one could renumber at will and multihome by connecting to as many
> providers as you were willing to pay.  However, looking at the current
> "operational experience" of which I am aware, neither of those
> deliverables has yet become real.  Are you aware otherwise?

Unfortunately, no.  And as I have indicated on this mailing list I'm 
not sure the IETF approach for multihoming in IPv6 (shim6) is a 
realistic solution to the need to multihome for the foreseeable future.

> Sticking with the "no PI space" mantra for IPv6 in the absence of 
> viable
> multihoming and/or easy renumbering is just plain wrong.  We need to 
> come
> up with a criteria for PI space that will constrain the recipient pool 
> to
> a reasonably small set of large institutions, i.e. those who will be
> multi-homing and for whom renumbering would be a major and painful 
> task.

It can, of course, be argued that renumbering networks that do not have 
a dedicated IT staff or significant technical expertise can also be a 
painful task (albeit not as major as a large enterprise).  In fact, I 
suspect the cost of this sort of renumbering will likely create a 
market for NATv6, but I know saying that is heretical...

To be clear, I am not sticking with any particular mantra, I was simply 
pointing out the reality of the situation.  It can be convincingly 
argued that in the lifetime of IPv6, technology will advance such that 
any number of prefixes that might be created could be handled, just as 
the routing table growth of IPv4 was handled (although I'd personally 
prefer it be handled a bit less ... stressfully).  It can also be 
convincingly argued that (to paraphrase Dave Clark) the IPv6 truck is 
driving into the same swamp as IPv4, yelling "me too, me too".

Simply, renumbering sucks.  Not being able to multi-home in a way that 
doesn't impact the DFZ sucks.  Creating a "pioneer's reward" can also 
suck in the inequities that are created, particularly when a limited 
resource (routing slots) becomes scarce (e.g., I personally experienced 
the implications of this inequity when trying to explain that despite 
the fact that (at the time) Stanford had a /8 was not sufficient 
justification for 20 /8s and 20,000 /16s for the entire country of 
China).

Everything has a cost.  The question isn't either/or, it is a question 
of where we draw lines.

> I invite all PPML readers to help us to reach that goal...

As do I.

Rgds,
-drc