IPv6 PI allocation

[Roger Jørgensen]

On lør, mai 19, 2007 23:41, Gert Doering wrote:
> On Sat, May 19, 2007 at 03:56:38PM +0200, Roger Jorgensen wrote:
>> in short, the only real option we have is IPv6.
>
> Do you have already started to plan this, and if yes, are you permitted
> to talk about it?

I got the attention from everyone at work after I presented the numbers
for when we´re out of IPv4 addresses and also told them the obvious
solution, IPv6. All feedback, including CEO and other managers was
positive so we´ve got a "go ahead" order to work more with IPv6. Hope to
have some real rought plans ready for presentation internaly within a few
months just to kick of some more official projects for moving along with
IPv6. It´s obvious for everyone as far as I know that we have to prepare
for IPv6, and also that we can gain alot from moving to IPv6.


> I'm mainly curious about the addresses you plan to use (ULA-local,
> ULA-central [which don't exist yet], RIR /32s, PI, ...) - and how well
> your corporate application vendors handle it.

That´s the real big problem. I´ve had several discussion with one of the
people that build the current network at work and we agree on one thing,
we need to learn from the mistakes done and make the next generation of
the network better. How is still an   unanswered question.

But for the  structure we´ve so far considered two models
a) We can go for ULA local/central combined with RIR space
b) or only RIR space

We just recently become a LIR to get our own IPv4 address space that we
control. PI space wasn´t an option as far as I know. I think we gain more
from being LIR than if we just requested address space. And for those that
want PI, become a LIR. It might cost more but you gain more in the long
run.

Anyway, the options,  a) save us some troubles but ULA-central part worry
me. We really need _global_ unique netblocks for all our usage. We just
don´t want to get into the troubles with have today with collision and NAT
both ways to get things working. Without ULA-central we will atleast have
a need for one single authority handing out all ULA addresses for all the
organization we interconnect with. Not sure that´s possible...

Options b) probably wont work either, we can easily satisfy the requirment
for a /32 after todays and any changes I´ve seen suggested so far. Optimal
we would love to have atleast two /32, that mean a /31 or bigger but that
required more documentation, not sure it´s worth the time really.  A /32
would work just fine if we go for option a) but that has it´s own problems
as mention...


About the vendor part... don´t know much about the network since I´m on
the system group and work with servers. But most of the network equipment
I know about are bought within the last 5 years so think most of it
support IPv6, how well is another question.
On the system side we mostly run Linux, *BSD and other flavours of *NIXes,
even a few odd windows boxes and they all support IPv6 more or less. Same
for the application/services we have, we just haven´t actived it anywhere.

Anyway, the problem is not on the network or the system side, its on the
end-user side. Most of the application they use and need are written for
some very specific tasks and I really doubt alot, if any support IPv6 at
all on that side...



-- 

------------------------------
Roger Jorgensen              | - ROJO9-RIPE  - RJ85P-NORID
roger at jorgensen.no           | - IPv6 is The Key!
-------------------------------------------------------