On Tue, Sep 02, 2008 at 10:05:29PM +0200, David Monosov wrote: Hello, BTW, your v6 MX is down :-)
To be honest, my intention to see 2007-01 put in action is to solve the imbalance imposed by the current billing scheme (PA has a recurring, for most networks not too small price tag attached, PI is basically free). Don't tell me PI is not PA because you cannot assign it to other endusers, there are dozens of ISPs that run entirely on PI. The proposal itself states that end users will be subject to a fee. Based on your point above I take it that you favor interpreting this is "a non-trivial fee", and it seems to me that you implicitly suggest it should be higher than the costs the RIPE NCC will incur directly from processing such end user assignments.
I was thinking of about 100 EUR/y. This is not community consensus or has been discussed anywhere, this is the amount of money I thought of a couple of years ago (before 2007-01) when thinking about a model for IPv6 PI. I have accepted that we need some sort of PI, but noone wants the same swamp "everyone who tells a few fairy tales gets free space without any contractual relation" space that was there with IPv4 already. So we discussed a model similar to a domain registrar, with recurring fee paid by the enduser. If it's not paid it expires. Of course it isn't that easy for IP addresses, as you cannot revoke an announcement another ASN is doing that easily. Basically the yearly invoice should start people thinking about whether they still need that resource. Maybe they've merged a decade ago and the only device still using the old address space is the old 9pin dot-matrix printer in the basement. Maybe something else that can be easily renumbered. Maybe it's free already. I've seen "networks" claiming to multihome that run two tunnels through their dynamic SOHO DSL line. Or it seems to be a sport for a number of people in this business to air their own prefix and even ASN, even though they only have their Xbox and their laptop in it and peer with their workplace and a colleague. They won't let go of this, but at least they should be reminded that eats a FIB slot everywhere. On the other hand, whoever _needs_ to be multihomed or _needs_ to avoid renumbering should be able to do that. Serious multihoming is more expensive than the number I had in mind (physical diversity, anyone?). If your network is sufficiently complex that renumbering it takes a long time (=is "expensive" work), then you can invest the money into PI to avoid it as well. And no, I don't believe in that whole "emerging market, we can't pay it, you're killing our industry and our democracy with it" thing that someone is probably going to raise (or at least think).
So basically, bill new resources, get a sensible approach with old resources and finally get PIv6 out of the door. and while I don't see this draft as an explicit condition to getting IPv6 out the door (although I do see the benefits of doing so) and its intention is clear to me,
I do. I will raise hell against any PIv6 proposal that does not include some serious contractual relation with the enduser, and nothing forces updated contact information better than a yearly invoice to be paid.
The trouble, however, lies in the details. For example, while Shane's take on the implementation of this proposal is that it will result in no cost or minimal cost for end users - your interpretation seems to be one of an opportunity to enforce something akin to a routing slot tax which would be sufficiently high to discourage waste; two quite conflicting points of view stemming from the implementation of the the very same policy document!
I don't think 100 EUR/y would be "sufficiently high to discourage (deliberate) waste". It should force people to think about it. Every year when they receive their invoice. Apart from the few hardcore people that think it's their goddamn right to have their appartment network in every FIB worldwide it's a price vs. gain thing for most. At the moment the price is very low, often not directly associated with the resource at all. Thus even a very low gain or no gain at all is enough to keep the resource unchanged (not necessarily "used"). I want endusers to start thinking about this. Again, this number is my personal private opinion, it hasn't really been discussed with anyone. But yes, people expect different things of 2007-01, as with every policy. This is what I'm personally expecting/hoping. But I could not tell whether I'm voicing the single dissenting opinion or people actually agree with my arguments. Bernhard