New ARIN ipv6 allocation policies

[Jeroen Massar]

Bjoern A. Zeeb wrote:
[..]
> 1) Why are we talking about 'details' like PI when it's not even
>    possible to get native IPv6 connectivity to an entire continent?
>    Is IPv6 really at the stage to be widely deployed for business
>    then?

There is only about 1 operator in .za if you want cabling to the rest of 
the world, as such even getting IPv4 there is difficult. This is the 
whole problem with the situation in Africa, add on top of that 
apparently some nice monopolies and other problems.

Good thing about IPv6 in Africa etc: at least folks can get address 
space, as much as they want. And do to the great work of many, IPv6 is 
already there, now it just needs to get to the end-users.

Note that most businesses can't care less about Africa/South America and 
other developing continents. Their business is in the US/Europe/Asia, as 
such there is where they need it and there is where it is available.

> 2) Let's assume ARIN will allocate a /32 for those PI blocks,
>    what will happen if the other RIRs do the same? Ever though
>    about that and the possible growth about your rib?
>    Considering that you can get a /19 v4 PI today easily you can
>    expect that for IPv6 something similar will happen sooner or
>    later once PI is available. Maybe not a /19 v6 PI but PI for
>    everything you may need/want it for. Think about a 2nd /32
>    from each RIR or even more.

So? Then you and everybody else on this planet will need to upgrade 
their routers (I guess vendors are very pleased with this ;)

Also, a /32 is 'only' 65536 /48's * 6 of those blocks, maybe 300k 
routes. The amount of routes though isn't the issue, the issue will 
become the updates. But just buy a bigger router ;)

> 3) Filterig routes received from an IPv6 upstream these days
>    with a strict filter one only accepts about 90% of the prefixes
>    (610 of 671 atm here) from the DFZv6 because the others are
>    outside the allocation range etc. You can have a look at a sample
>    documented cisco prefix-list from 6/6/6 at [1] in case you accept
>    more prefixes;-) Will those more specifics go away once we'd have
>    PI or will people start announcing more and more /35 if we already
>    have more than enough of /48s?

The idea here is that PA is PA, and PI is PI. That means that the PA 
blocks _should_ only be announced as the full PA block and no more 
specifics. Which is why it is good that the PI prefixes come from the 
same block as then people who filter can easily filter them out; that is 
they can open their filters to accept the PI space.

In the end the internet is only for the folks with money: they can buy 
the equipment to actually use it or they can pursuade others to accept 
the prefix they want to use.

Btw, see http://www.sixxs.net/tools/grh/ for an overview of routes that 
are currently floating around on the internet.

Greets,
  Jeroen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 311 bytes
Desc: OpenPGP digital signature
Url : http://lists.cluenet.de/pipermail/ipv6-ops/attachments/20060905/c4c8fb21/signature.bin