Mikael Abrahamsson wrote:
On Tue, 15 Jan 2008, Marcin Gondek wrote:
- 100E yearly maintenance
If not higher. Hidden cost to the community is much much higher as each PI announcement translates (in the long run) to more expensive routing platforms.
We made a boo-boo with IPv4 in the PI area, let's not do it again with IPv6. Assign real cost to having a slot in the global DFZ. That way people that really really need PI will pay it (at 1000E a year it's still a win situation at a few tens of work hours) and the ones who do not, will not clutter the DFZ.
I second this (and the /48) suggestion ... from our experience, some customers who will most likely NEVER get a second uplink (don't really need it), and have very few boxes to change (in case of PA and provider change) have insisted to us to get PI instead of PA. No way of talking them out of it. Maybe with even a small _reasonable_ price tag for yearly renewal would have made it easier to reason with them ;) result is less unnecessary routing table growth for BS applications, and (also important) return of unused (read: unpaid) PI to RIR. In essence, here's my "wish list" for this topic: - /48, not /56 - small yearly fee (100-200€?) for routable PI - small one-time fee (100-200€?) for non-routable PI (take them out of a defined /32 or so which is/can/should be filtered by ISPs) - don't automatically distribute IPv6 space (both PI and LIRs) - e.g., our customers with PI I don't see using IPv6 within at least the next 5 years (we've had v6 available for several years ...). Instead, make initial v6 applications and allocations "painless", allowing for any PI owners to just "lift their hand" and they receive their v6-PI (/48) for just paying ;) -garry