Re: Sprint's route filters and Europe
We charge *everyone* for registration services. That is how it should be. There is no reason why governments (read: taxpayers) should be footing the bill.
This would be ok if there was a market for registry's. The current Ripe approach seems to build a large byracrazy, with lots and lots of paperworks, so you need more resources to process the paperworks, does not really support the part of the Internet we are where supposed to facilitate. If there is a finite resource that needs to be managed, it should be done in a fair way for all players, and right now i don't think that's not the case if we look at the globe as a whole. Fix the rules, open up more registrys wich all apply the same rules, maybe someone needs to audit the registrys to make sure that things are done correct, and have them compete on speed and effectiveness and cost. --Peter
Peter, [IEPG added as this isn't just a North American issue] Executive summary: come to the PIARA and IRE BOFs at the Montreal IETF.
We charge *everyone* for registration services. That is how it should be. There is no reason why governments (read: taxpayers) should be footing the bill. This would be ok if there was a market for registry's.
While I am generally a strong proponent of creating a market for registry services, I will be honest and admit things just aren't set up for the registries to compete yet.
The current Ripe approach seems to build a large byracrazy, with lots and lots of paperworks, so you need more resources to process the paperworks, does not really support the part of the Internet we are where supposed to facilitate.
From my (admittedly biased) perspective, it would seem there are two
The existance of a bureaucracy and "lots and lots of paperworks" is the direct result of what I consider the complete raving irrationality of the current registry system. What would you propose the registries do to meet the requirements of a) conserving address space, b) conserving routing table space, and c) allocating the remaining free pool of addresses in a "fair" fashion when we're constrained by (arguably) obsolete policies put in place in the days when the Internet was one big happy R&E community? options: A) The socalist approach B) The capitalist approach Right now, the registries use option A. Addresses are allocated "to each according to need" and the regional registries that charge (APNIC and RIPE-NCC) request "from each according to ability" (at least to some extent). As a result, since the definition of "need" is hard to pin down, a bureaucracy is created and (since there is no other acceptible way of verifying needs and/or reducing the address allocation rate), huge amounts of paperwork and oodles of tedious forms are necessary to submit requests. Every time someone (who me?) brings up option B, we go chasing merrily down one or more of the following ratholes: 1) we need to conserve route table space, lets charge for that, not addresses (irrelevant) 2) AT&T (or some other evil speculator) will buy up all the address space (and ISPs are just going to sit idly by?) 3) if you charge, then poor organizations can't connect to the Internet (so who's paying for their connectivity?) 4) you can't charge for addresses because they're just numbers and have no value (tell that to the US Treasury) Regardless of the validity of any of these ratholes, they are missing the point -- without some OBJECTIVELY VERIFIABLE CRITERIA, the registries must rely on people being honest and forthright about their requirements. Money (e.g., a charge per address) has proven to be a pretty effective objectively verifiable criteria to determine whether someone *really* needs the address space they are requesting. However, since option B is theo-politically infeasible for whatever reason, you get option A, with increasingly draconian rules and ever growing mountains of paperwork. Life is harsh.
If there is a finite resource that needs to be managed, it should be done in a fair way for all players, and right now i don't think that's not the case if we look at the globe as a whole.
What is "fair"? Is it fair early adopters have (mutliples of) /8s and will never need to go through the registry hassle? Is it fair that the current allocation policies are (statistically speaking) conserving address space to the benefit of the large ISPs, most of which are in the US? Would it be fairer if the registries allocated /14s (or /19s) to everyone regardless of requirements? Should everyone who wants an IP address be given one, regardless of what they'll use it for? Should addresses only be given to ISPs? The registries try to be "fair", from their perspective. Presumably, the concern you are expressing here is that the regional registries have implemented the various policies somewhat differently. I believe this was a specific goal of RFC 1466 which created the multiple registries in the first place. The registries are trying to coordinate policies, but we're somewhat constrained by the different communities we serve (e.g., what is considered a "small ISP" in the US is likely to be larger than most of the ISPs in the AP region).
Fix the rules,
Not to pick on you in particular, but I didn't see a lot of comment from you or other people on the registry guidelines draft. Please indicate what rules need to be fixed (I have my own set, but I'd like to see other people's).
open up more registrys wich all apply the same rules,
Applying all the same rules would tend to imply ignoring the differences in the development in the Internet throughout the world. Is this what people really want (honest question -- there are arguments on both sides)? Oh yeah, check out how many bits there are for registries in IPv6...
maybe someone needs to audit the registrys to make sure that things are done correct,
Who and who would pay them?
and have them compete on speed and effectiveness and cost.
I think we agree this should be the end goal. How do we get there? For those who are interested in this gunk, there will be 2 BOFs at the Montreal IETF: Pricing of Internet Addresses and Route Advertisements (PIARA) and Internet Registry Evolution (IRE) I'm sure both BOFs will be non-controversial and quite sedate. Stop on by if you want to catch up on your sleep... :-). Regards, -drc P.S. So which rathole are we going to go down this time? 30 quatloos on rathole #4 (it's my favorite 'cause it's so silly (Hi Brian!)).
4) you can't charge for addresses because they're just numbers and have no value (tell that to the US Treasury) ... P.S. So which rathole are we going to go down this time? 30 quatloos on rathole #4 (it's my favorite 'cause it's so silly (Hi Brian!)).
Well, since you ask... My view is still that charging for routing announcements is necessary and sufficient - but as I've said before, I think that registries should charge for their *services*, and allocating an address block is a service. So I don't think we're that far apart. See you in Montreal. Brian
Brian,
My view is still that charging for routing announcements is necessary and sufficient -
Sufficient for what?
but as I've said before, I think that registries should charge for their *services*, and allocating an address block is a service.
What is the difference in the service charge for allocating a /8 compared to allocating a /24?
So I don't think we're that far apart.
You're in Switzerland, I'm in Japan... Cheers, -drc
Randy,
My view is still that charging for routing announcements is necessary and sufficient -
Sufficient for what?
Sufficient to give folks an incentive to use addresses that aggregate for routing purposes. (This is carefully phrased *not* to be CIDR-specific.)
but as I've said before, I think that registries should charge for their *services*, and allocating an address block is a service.
What is the difference in the service charge for allocating a /8 compared to allocating a /24?
Indeed - I was debating whether to mention that in my previous mail. I don't have a quick answer. Brian
From my (admittedly biased) perspective, it would seem there are two options:
A) The socalist approach B) The capitalist approach
The argument appears to be more well-rehearsed than that. In essence it seems similar to the name-space argument. We have a scarce resource, and must find ways of distributing it. Past experience has shown that the free market approach is in general the least of all evils, but it has some pathologies; these are well known in economics in general not just in terms of internet politics. The main issue is that we have a single supply (noone can go and set up another IP address space), and the cost price is apparently zero. For instance:
1) we need to conserve route table space, lets charge for that, not addresses (irrelevant)
Specific case of peculiar non-linear cost curve complicated by the fact that there has to be some economic disadvantage to poor aggregation on a given amount of address space.
2) AT&T (or some other evil speculator) will buy up all the address space (and ISPs are just going to sit idly by?)
Specific case of monopolistic disfunctionality.
3) if you charge, then poor organizations can't connect to the Internet (so who's paying for their connectivity?)
Specific case of the merit good argument.
4) you can't charge for addresses because they're just numbers and have no value (tell that to the US Treasury)
Urmm... see the market for options and derivatives. Historically the way to prevent such market disfunction has been regulation of this sort. Which is exactly what Internic, RIPE, etc. do. However, where regulation is different in different geographical areas in what is effectively a global market place, it causes problems. c.f. the telecoms industry. Alex Bligh Xara Networks
"Alex.Bligh" <amb@xara.net> writes:
Historically the way to prevent such market disfunction has been regulation of this sort. Which is exactly what Internic, RIPE, etc. do. However, where regulation is different in different geographical areas in what is effectively a global market place, it causes problems. c.f. the telecoms industry.
That is true and we work very hard to align all regional registries to the extent possible. I see no major differences causingproblems at the moment. If there are, please point them out. randy Conrad has made the argument why there is a need for the regional differences that do exist. Daniel
Daniel Karrenberg <Daniel.Karrenberg@ripe.net> wrote
"Alex.Bligh" <amb@xara.net> writes:
Historically the way to prevent such market disfunction has been regulation of this sort. Which is exactly what Internic, RIPE, etc. do. However, where regulation is different in different geographical areas in what is effectively a global market place, it causes problems. c.f. the telecoms industry.
That is true and we work very hard to align all regional registries to the extent possible. I see no major differences causingproblems at the moment. If there are, please point them out.
randy Conrad has made the argument why there is a need for the regional differences that do exist.
I don't disagree with either of the above. I was not arguing against regional registries, just that the problem is more widely known and understood than in just the internet industry. There has been the potential for problems (viz. the subject line) but AFAIK they've mostly been sorted by harmonizing the differences. Where this hasn't happened yet (one area I have direct experience of is namespace - .co.uk vs. .com for example), it has started to distort the 'market'. Alex Bligh Xara Networks
"Alex.Bligh" <amb@xara.net> writes:
Historically the way to prevent such market disfunction has been regulation of this sort. Which is exactly what Internic, RIPE, etc. do. However, where regulation is different in different geographical areas in what is effectively a global market place, it causes problems. c.f. the telecoms industry.
That is true and we work very hard to align all regional registries to the extent possible. I see no major differences causingproblems at the moment. If there are, please point them out.
I agree. The regional registries meet several times a year as well as keeping in close contact via email to coordinate our policies as much as possible. We are aware of geographical differences and try to make allowances for them as much as possible. I think the majority of the problems with IP numbers pretty much transcend any international boundaries. Kim
Daniel
A) The socalist approach B) The capitalist approach The argument appears to be more well-rehearsed than that.
"Well-rehearsed"?
In essence it seems similar to the name-space argument.
If you mean the domain name-space, then I would disagree. For all intents and purposes, the domain name space can be considered infinite.
We have a scarce resource, and must find ways of distributing it.
Agreed. It is a standard "tragedy of the commons" problem with the same solutions.
Past experience has shown that the free market approach is in general the least of all evils, but it has some pathologies; these are well known in economics in general not just in terms of internet politics.
Yes, you are moving the problem around. However, people seem to prefer the ability to pay for something and be done with it over filling out forms and waiting in (virtual) lines.
The main issue is that we have a single supply (noone can go and set up another IP address space), and the cost price is apparently zero.
How much does water cost?
For instance:
1) we need to conserve route table space, lets charge for that, not addresses (irrelevant)
Specific case of peculiar non-linear cost curve complicated by the fact that there has to be some economic disadvantage to poor aggregation on a given amount of address space.
2) AT&T (or some other evil speculator) will buy up all the address space (and ISPs are just going to sit idly by?) Specific case of monopolistic disfunctionality.
The creation of a monopoly in IPv4 address space can't happen. Too much has been distributed to too many players. Further, even assuming somebody goes out an buys all the address space, as soon as they try to put the screws to MCI, Sprint, UUNet, etc., those service providers would likely simply declare some chunk of the address space as "theirs" and start routing it. The only reason the service providers do not do that now is because the registries provide a neutral, universally accepted forum in which address validity is defined. Without the registries (e.g., in the case of a monopoly or cartel), address validity would need to be negotiated among all relevant players. As the definition of "relevant" is somewhat nebulous, it would seem unlikely everyone would agree.
3) if you charge, then poor organizations can't connect to the Internet (so who's paying for their connectivity?) Specific case of the merit good argument.
Good argument? If an organization wants to connect to the Internet, it is almost certain the address cost will be so far down in the noise as to be undetectable. If address costs were a significant portion of the total costs, there is always NAT/ALG.
4) you can't charge for addresses because they're just numbers and have no value (tell that to the US Treasury) Urmm... see the market for options and derivatives.
Agreed.
Historically the way to prevent such market disfunction has been regulation of this sort.
What dysfunction are you trying to prevent?
Which is exactly what Internic, RIPE, etc. do.
No, that might be what the could do, but it isn't what they do now.
However, where regulation is different in different geographical areas in what is effectively a global market place, it causes problems. c.f. the telecoms industry.
Agreed. That's why the regional registries try to work together. Regards, -drc
Try again. exmh ate the last one. David R. Conrad" <davidc@apnic.net> wrote:
A) The socalist approach B) The capitalist approach The argument appears to be more well-rehearsed than that.
"Well-rehearsed"?
= economists have covered a more general case of this argument many times before.
In essence it seems similar to the name-space argument.
If you mean the domain name-space, then I would disagree. For all intents and purposes, the domain name space can be considered infinite.
Domain space is different in that varies considerably in usefulness. Though the whole space is infinite, the useful parts aren't. For instance names which do, or might, represent company/corporate names are much more valued than those that don't. Having just spent a long time looking into how we are going to change the UK namespace around, this is definitely the prevailing opinion here anyway. Recent litigation in the US seems to imply the same thing applies to .com.
The main issue is that we have a single supply (noone can go and set up another IP address space), and the cost price is apparently zero.
How much does water cost?
To consumers in the UK? Quite a lot. I guess you are making the point (which I'd agree with) in that the cost is in the infrastructure. So the water companies add value to a 'free' raw material, and then charge for nice clean water delivered in a pipe. The difference here is that it's the registry who gets paid, but the large ISPs who pay for the infrastructure, so the economics gets skewed. AFAIK the registries don't fund the ISPs proportionately to how many prefixes they give out. Indeed to some extent the reverse is true.
3) if you charge, then poor organizations can't connect to the Internet (so who's paying for their connectivity?) Specific case of the merit good argument.
Good argument? If an organization wants to connect to the Internet, it is almost certain the address cost will be so far down in the noise as to be undetectable. If address costs were a significant portion of the total costs, there is always NAT/ALG.
Crossed wires. I meant the "merit good" argument. Most countries still agree, for instance, that funding healthcare is a good thing to varying extents: by subsidizing treatment of poorer people's infectious diseases, those who can aford to pay are less likely to get them themselves. Thus we have little cholera in the west. If we believe that subsidizing "poor organizations" connectivity to the internet is (economically) in all our interest, then there is an economic justification for subsidy by the "merit good" argument. We might, for instance, feel that not getting poorer sections of society connected is likely to disenfrachize them. Or we might not.
Historically the way to prevent such market disfunction has been regulation of this sort.
What dysfunction are you trying to prevent?
Suboptimal distribution - i.e. lack of a pareto-optimum. The price mechanism only produces pareto-optimality (i.e. the best result) under certain circumstances. Monopoly is one dysfunction that will stop it doing so. As, I gues, is the wierd cost curve we have. However my economics is too rusty to prove it, though it would probably make a fair thesis for someone.
Which is exactly what Internic, RIPE, etc. do.
No, that might be what the could do, but it isn't what they do now.
They replace the price mechanism with an imposed system of distribution. Whether you like it or not is another argument. I think they do a fair job. Alex Bligh Xara Networks
David R. Conrad sez:
3) if you charge, then poor organizations can't connect to the Internet (so who's paying for their connectivity?) Specific case of the merit good argument.
Good argument? If an organization wants to connect to the Internet, it is almost certain the address cost will be so far down in the noise as to be undetectable. If address costs were a significant portion of the total costs, there is always NAT/ALG.
I am sure that there are enough mature technology solutions to ensure that "poor" organisations can still use the Internet effectively. The bottom line is that addresses **are** a scarce resource. The APNIC has been encouraging ISPs in the region to do sub-C allocations and that will reap its own benefits despite the fact that there is an overhead in maintenance. There is also a substantial amount of user education that is required in this part of the world to convey the fact that it is not **necessary** that all the hosts in your organisation be on the Internet. (I wouldn't see this statement as out of context here, since the discussion is all about charging, anyway ..:) ) In quite a few ways, the AP region is distinct in its approach to the Internet. BTW, cld I ask if the GISD document is available in any complete form ? Thanks --Gopi Garge ERNET, INDIA
gopi@ece.iisc.ernet.in (Gopi K Garge) writes:
BTW, cld I ask if the GISD document is available in any complete form ?
The framework is there. There was little impetus by anyone to contribute to it afterwards. The sentiment I got from many people was 'Why publish a handbook for the competition?'. Once that was established the project had served its purpose. Daniel
participants (7)
-
Alex.Bligh -
Brian Carpenter CERN-CN -
Daniel Karrenberg -
David R. Conrad -
gopiļ¼ ece.iisc.ernet.in -
Kim Hubbard -
Peter Lothberg