From: Nick Williams <nmw@haven.ios.com> If a registries' actions or lack thereof hurt someone else's business they may end up being liable in court for it Oh, great, let's put the registries out of business. I mean, who needs the aggravation, and the threat of being sued? I urge registries to consider charging for their address space delegation services This is absolutely not the solution. This still does nothing to slow the growth of the routing tables, which is the *real problem*, &^%!@$*#&^%! I mean, how is it going to stop me moving from provider A to provider B and wanting to take my addresses with me if I had to pay for the addresses? The only charging scheme that I can see working is to charge people for advertising routes, with the charge related to the scope over which the advertisement is seen. That way, there's a direct relationship between the amount of resources consumed, and the amount charged, which is utterly fair. Note that if you have a "fully" "portable" address (i.e. one advertised over a global scope), your monthly bill will go up as the Internet gets larger, under this scheme... If registries can make decent allocation decisions ... then I argue that routing table growth will be curbed and that IPv4 address space utilization effeciency will rise. First, you're talking about two completely separate problems. Let's keep address space utilization out of it for now, OK? Second, even if registries did allocate addresses optimally, what happens when those sites move around, something the registries have no control over? > "Sorry, we are only able to provide you with partial Internet service > at the moment because Sprint doesn't like the addresses we assigned you". It is costly, as the solution is to become a client of Sprint. So, everyone, can we find a solution? That may solve it for Sprint, but suppose you have the same problem with otehr major providers? Do you have to get a link to each of them? When does this cease to become distinguishable (costwise) for charging for routes - actually, the latter would probably be less expensive. Noel
Noel, The only charging scheme that I can see working is to charge people for advertising routes, with the charge related to the scope over which the advertisement is seen. That way, there's a direct relationship between the amount of resources consumed, and the amount charged, which is utterly fair. Since we can now talk about charging: A global charging scheme implies that there is a centralized, global charging authority. That's unlikely to fly in any case. What MAY fly, however, is if individual backbones charge for advertisements as well as bandwidth. This would be charged back through regional ISPs, who would decide which backbones to connect to and charge back to their respective customers. Tony
From: Nick Williams <nmw@haven.ios.com>
If a registries' actions or lack thereof hurt someone else's business they may end up being liable in court for it
Oh, great, let's put the registries out of business. I mean, who needs the aggravation, and the threat of being sued?
I'm not arguing in favor of anyone suing any registry, nor am I threatening to do so myself; I'm merely pointing out that some fellow out there might have the reasons, resources and incentives to go after registries in court.
I urge registries to consider charging for their address space delegation services
This is absolutely not the solution. This still does nothing to slow the growth of the routing tables, which is the *real problem*, &^%!@$*#&^%! I mean, how is it going to stop me moving from provider A to provider B and wanting to take my addresses with me if I had to pay for the addresses?
The only charging scheme that I can see working is to charge people for advertising routes, with the charge related to the scope over which the advertisement is seen. That way, there's a direct relationship between the amount of resources consumed, and the amount charged, which is utterly fair.
Routes have two costs: they represent IP address space (of which there is a shortage right now) which is used inefficiently, and they use up router resources in a menner that's inversely proportional to the amount of IP address space the route represents. Hence you have two opposing costs. I'm arguing that economics have to enter the picture for any application for IP address space a registry receives. ISP/NSPs can deal with some of theis on their own, but registries are at the root of all IP address allocations, so therefore they are a good place to look towards for an injection of economic realities into a process that so far has seemed inexpensive: obtaining usable IP address space. Unfortunately you cannot retroactively, unilaterally amend the many existing contracts between ISPs, NSPs and customers, so we have to look to charging new entrants only, and old participants when their contracts come up for renewal. My goal in asking that the registries charge for their services is: better services, better allocations of IP address space (i.e. allotion of IP address space to those who are serious about using it).
Note that if you have a "fully" "portable" address (i.e. one advertised over a global scope), your monthly bill will go up as the Internet gets larger, under this scheme...
Sure, but such a scheme cannot be implemented right now. How is an organization going to figure out that one particular /24 can be charged while another can't be?
If registries can make decent allocation decisions ... then I argue that routing table growth will be curbed and that IPv4 address space utilization effeciency will rise.
First, you're talking about two completely separate problems. Let's keep address space utilization out of it for now, OK?
But the registries don't seem willing to stop considering IP address space exhaustion as a BIG problem, and I don't blame them. We have two problems which are interrelated; you can't ignore one and solve the other, because the one you ignore may become far worse as a result of your solution.
Second, even if registries did allocate addresses optimally, what happens when those sites move around, something the registries have no control over?
Ah, well, these are new customers to other NSPs, so NSPs can refuse to take their CIDR holes elsewhere. If these new entrants go directlry to NAPs, then they are not likely to move, but if they do, then they may have a hard time getting peers to take their long CIDR holes. Then again, folk who go to the NAPs end up having few small blocks in the long run, because they tend to be successful (I hope, maybe I'm wrong?) and so the registries have less misgivings about giving these /16s rather than /20s.
"Sorry, we are only able to provide you with partial Internet service at the moment because Sprint doesn't like the addresses we assigned you".
It is costly, as the solution is to become a client of Sprint. So, everyone, can we find a solution?
That may solve it for Sprint, but suppose you have the same problem with otehr major providers? Do you have to get a link to each of them? When does this cease to become distinguishable (costwise) for charging for routes - actually, the latter would probably be less expensive.
You're right, charging for route announcements is something that has to happen. I've argued this myself too. But what about the registries worries wrt IP address space exhaustion?
Noel
Nick
I'm arguing that economics have to enter the picture for any application for IP address space a registry receives.
As I have mentioned before, in the case of APNIC, there will be a cost associated with requesting address space: US $10,000/year for Large ISPs US $5,000/year for Medium ISPs US $2,500/year for Small ISPs US $1,500/year for Non-ISPs US $1,000 startup fee. The category is self-determined by the organization. The charging scheme used by APNIC is nearly identical to that used by the European registry, RIPE-NCC, varying only in the amount charged (last I checked). Obviously, a college student with four modems in her garage is most likely not going to be able to afford US $2,500/year for registry services, thus it is conceivable that ISPs in this class will obtain their address space from their upstream provider. Arguably the correct place. Further, non-ISP organizations will be obligated to pay US $2,500 when they request the space, US $1,500 per year for continued APNIC services, so it is likely the small (/24+) requests will tend to move towards their providers which (presumably) are significantly cheaper. Again, arguably the correct thing happens. Such an approach is perhaps not as clean as charging for IP address space directly, but it does seem to be a reasonable compromise given political and theological constraints. Note that an organization claiming to be an ISP will obtain a block of sufficient size to avoid most proposed prefix length filters. However, at this time, I am unaware of any plans InterNIC has regarding charging for IP registration services. InterNIC's approach seems to be (speaking from observation only) to request "small" ISPs to obtain their space from their upstream provider, however if the ISP insists, InterNIC will allocate them long prefix provider independent space (as exemplified by the prefix that started this thread).
My goal in asking that the registries charge for their services is: better services, better allocations of IP address space (i.e. allotion of IP address space to those who are serious about using it).
The place you should be directing such comments to is the US NSF. Note that you should stress charging for registration services as opposed to charing for address space. The latter is a political minefield that will simply muddy the waters.
But the registries don't seem willing to stop considering IP address space exhaustion as a BIG problem, and I don't blame them. We have two problems which are interrelated; you can't ignore one and solve the other, because the one you ignore may become far worse as a result of your solution.
As I pointed out a couple of days ago, there is a high likelihood that to encourage CIDR, you will discourage address conservation. However, my observation is that there is a general feeling that address space exhaustion will take care of itself -- as the address space becomes less available, people will: a) become more efficient in utilizing the space b) use NAT/ALGs c) obtain address space in "non-traditional" means (e.g., buying it from another organization (how many As does BBN have? :-)). d) migrate to IPv6 See RFC 1744 for more details. Note that option c is theoretically not available -- according to current and likely future guidelines, buying/selling of addresses is essentially disallowed, addresses should be returned to the orgininating registry to be reallocated as criteria are met. I would be very interested in hearing people's opinions regarding this aspect of current allocation guidelines. Regards, -drc
participants (4)
-
David Conrad -
jnc＠ginger.lcs.mit.edu -
Nick Williams -
Tony Li