______________________________________________________________________ DRAFT TRANSCRIPT Session: APNIC Member Meeting Date: Friday 3 March 2006 Time: 9.00am Presentation: APNIC fee structure Presenter: Paul Wilson ______________________________________________________________________ PAUL WILSON: The issue of the APNIC fee structure was mentioned. We did have a session, an open session about the fee structure during the week, and that was one which was attended by a few people, but certainly not as many as we have in the room here. So what I would like to do is to make a similar presentation here, as I did at that open session, to prompt a little more thinking and a little more awareness of the issues we're facing of the APNIC fee structure and the solutions we may have. This is not a proposal for a fee structure, what I'm doing here is describing the framework we have for the free structure today, and how we might adjust the framework as an example to achieve some of the solutions, or the objects that we seem to have with the fee structure today. So I'd ask you to bear that in mind as I make this presentation. It is an example of a fee structure rather than a specific proposal. In order to turn this proposal into an example, I think there would have to be more discussion and there will be a mailing list established, if not already, which will be announced and which will allow the issue to be discussed as time goes on by APNIC members. OK, what I'm going to do is talk about a bit of a background for this problem. Some of the principles on which the fee structure needs to be or is established. This example for a candidate fee structure solution and how we were to assess its impact, some implementation issues and maybe some discussion about it. One of the background issues has got to do with APNIC's financial position and how that position might change in the future. What this chart shows is that over a number of years between '99 and 2002, APNIC enjoyed some quite substantial surpluses, revenue substantially higher than expenses, which was very fortunate, because we were growing very rapidly at that time. And the EC has a decision in place that APNIC must maintain a capital reserve, sufficient to the next year's budget. When your next year's budget is growing that rapidly, you need a surplus in order to keep that capital reserve up to equal the following year's budget, and that's what we had in those years. And that was quite useful. From 2003 and onwards, we've had balanced budgets and a very balanced situation in 2004. 2005 was projected to be a lineball budget which revenues equalled expenses and we only had a surplus because of the foreign exchange gains, which I mentioned earlier. In this year, we have a budgeted loss, a small loss, which we'll be aiming to overcome, if we can, depending on circumstances. But that's what the projection is at the moment. The point here is that it looks good for APNIC to have a balanced budget. We're not aiming for its own sake to make a profit or make a surplus each year, but what I think you saw during the presentation of the annual report is that APNIC services are increasing in line of demands from the community for preferred services and they're increasing in line with demands that we don't seem to be able to do much about, such as these, let's say, political pressures that have been applied and seem to be increasing at the moment. They came about - firstly, I have to say through the whole ICANN process, and since then compounded by the WSIS and related processes. We're incurring larger overheads of that nature, and in some sense we also have to be careful about what overheads we might incur in future. We have to be in a stable and secure position in order that some change in future doesn't adversely affect and suddenly affect the viability of the registry. We do have a one-year capital reserve but no-one likes to eat into that unless it's necessary and we may well eat into that this year. But having started on that process, possibly going into some deficit in the annual budget, it is time to have a look at whether we're actually doing the right thing financially with the fee structure and so forth. There's a list here of the new activities in recent years, this is not just over the last year which you've heard about, but over the last few years, during which, as I say, we've had balanced budgets, but we have managed to increase activities, services and, of course, expenses during that time. I won't go into all of these, I think you see what they are - it's got to do with rootservers which we've been asked to deploy some very practical and important research and development actions, activities, training, which is under increasing demand, meetings, including increased participation in meetings. And so on and so forth. So these things will continue in future, at least the APNIC EC so far has expressed a comfort with the activity levels that we're undertaking and the budgeting and so on. The future needs to be a little bit more secure than possibly it is at the moment. The current fee structure we have was established in 1996, 10 years ago. At that time we had three tiers of membership - small, medium and large. And we had the same annual fees as we do today, $25,000, $10,000 per annum. The fee structure was optional. You decided how big you wanted to be. You could select to be small or large. That worked quite well while the APNIC community was small and it worked well until the Asian economic crisis of about 1997- '98. But at that time the option to select a small membership tier was too attractive or irresistible for companies that may have had financial worries. It was quite a few years ago in 1999 that we established the mandatory tier selection. Small, medium or large was determined by the address holdings and that is not meant to be a case of paying for addresses, but the address holdings do seem to give a reasonable indication of the size of a company, the capacity of a company to pay to support APNIC's operation. And possibly it seems to be a reasonable way in which to base the annual fee. So we have added additional tiers in the last few years to the APNIC fee structure. But, in fact, we haven't inflated annual fees for any of the inflation that's gone on over the years, nor for the fact that the value of the US dollar has reduced. So the reducing value of the US dollar is quite useful for members, in that the fees that you pay in real terms in your own currency tend to go down. It's the opposite to the APNIC Secretariat because most of the income of APNIC is in US dollars and most of the expenses are not in US dollars. In fact, we've suffered a double blow, if you like, in that there's been a lot of inflation over the last 10 years, certainly, and over the last six or so years, this devaluation which has reduced the value of the US dollar for the Secretariat. What does the current member fee structure look like? On an annual assessment of the address space holdings of each member. And the total address space holdings are calculated to a prefix, a particular prefix, in either IPv4 or IPv6 and there is a prefix range that sets your membership tier. In the table here - (refers to slide) - you can see the small membership level is $2,500 a year. If you have up to and including a /19, you'll be established at that small level. Once you exceed a /19 of IPv4 space, you'd be bumped up to a medium level and pay twice as much. It goes from 20 to 19 and 16. It's a factor of 8 in terms of the address space holdings. But a factor of 2 in terms of the annual fee increase from tier to tier. There are - the other aspect of the current APNIC fee structure, which we've heard a little bit about from Maemura-san, is the NIR fees which have been quite an consistent thing with the APNIC structure. So that's quite different from the annual fees which normal members pay. In the case of NIR, a per-address fee is paid at the time of allocation and only once only. So it can be quite a high fee in the case of large allocations. It's paid once, it's not paid again. What that means is that the situation for an ISP that is a member of an NIR is completely different and very hard to compare with the situation of an ISP that's a direct APNIC member. It brings up issues of fairness and consistency and also reliability of income, because, of course, APNIC has got no idea and the NIRs probably have not much more of an idea when these large allocations are going to occur. So you can have unexpected revenues or an unexpected lack of revenues depending on when the allocations are recurring. There's no actual fee charged for each member of an NIR but the NIR pays a total annual fee according to the same scale. Most of the NIRs are in an extra category because collectively they've received a great number of addresses. They're not assessed or calculated in terms of the number of ISPs actually involved, which can be in the hundreds in some cases. OK, so if we're going to have a look at revision of the fee structure, it's useful to ask why and what would be the rationale for changing the fee structure and what rationale, a new fee structure could be based? There's a question of overall revenue maintenance for APNIC itself. Part of that is the unpredictability of the current structure involving NIRs which is a problem for both APNIC, Secretariat, APNIC and the NIRs. There is a possibility of NIRs being formed almost at any time. And the impact of forming an NIR in a country like Australia or India would be pretty substantial in terms of migration of members from APNIC to the NIR itself. There's another rationale, as I've explained, is that we don't have any, we haven't heard any provision over the last 10 years for the effects of inflation or foreign exchange. There is a possibility where a revision of the APNIC fee structure to adjust the overall revenue levels. I'm going to suggest a fee structure that could be applied to the APNIC fee structure shortly and that fee structure is engineered and designed to provide an increase of between 10%-15% over the current revenues from the current fee structure. One of the underlying features of our fee structure is to base annual fees on address holdings and that's something that's not questioned at the moment. I don't know if anyone has ever questioned that since that was established, but the new fee structure or any example of fee structure that I show here is based on the same principle, reflecting the size of the organisation's capacity to pay a share of the APNIC operating costs. There's an issue in the current structure of small versus large members. As I said before, there's an 8-fold increase in address space holdings for each 2-fold increase in the annual fee and that over the whole range of tiers in the fee structure gives a pretty phenomenal difference in what you could calculate as the effective per-address case per year for those members. An ISP that has a /8 has got 16 million addresses, paying an annual fee of $40,000, which gives an effective per address cost to that ISP, off 0.2 cents per address per year. An ISP that charges you $10 or $20 a month for your dedicated address is doing pretty well on a 0.2 cent cost per address. The ISP that holds a /20 has only got 4,096 addresses, paying $2, 500 a year, they're paying 61 cents per address. Something that has been raised is an issue of fairness within the structure. An issue of fairness and consistency at least in the situation of a member of an NIR or a member of APNIC. And it's very hard to compare what is the cost and benefit, for instance, in each of those cases. So there are issues of fairness there as well. Also, under this question of rationale, it does seem that the fee structure to which NIRs are subjected should be similar to or comparable of the fee structure of normal APNIC members. That's in the interests of predictability and fairness. We do have a migration of members to and from NIRs because where an NIR is operating in an economy we do allow ISPs to choose one or the other. And it makes sense that in making that choice, there should be some easy comparison between the two options. So, in what I'm assuming in this current discussion is that any new fee structure would be based for members of NIRs on the APNIC fee structure. With an agreed discount, in recognition of the fact that NIRs provide real and valuable services to ISP and their economy. There's something which has been raised also a number of times is that the members of NIRs don't vote. An NIR which is assessed at the extra large level has 64 votes in the current structure. They might have 200 or 300 members and that entire membership body is related to or is represented by a total of 64 votes, which is just the same as any other extra large member of APNIC and proportionally a very small voting in total. OK, so just to recap on how the fee structure works. We have a framework where there's a minimum fee level and a minimum recognised address space holdings and that represents the lowest tier of the fee structure. And you increment your tier according to your address space holdings and you have an increase in the fee structure for each tier. The minimum fee level for address holders is $1,250. The increment is to double the fee for each tier. The address space at that level is anything up to a /22 and the incremented address space is three lifts per tier. It produces this table of fees. So, by way of example, we could type some different parameters into this formula, setting a minimum fee level of $360 - a fee increment again of double the tier, starting at the minimum tier of /24. That's quite useful because we have quite a number of members holding historical address space and a member holding a /24 currently, a $1,250 a year. And understandably, anyone who considers that as a fee that's payable for something that historically they previously received for free is not particularly happy generally to pay that fee. Now, from that lowest tier, an address space increment of two per tier is used in this example. And so using the space spreadsheet, we get a number of tiers. We're not going for in this example small, extra small, extremely small, but just tiers, A, B, C, D, E, F. A tier A is up to a /24 and costs $360 US a year. If that ISP is a /16, then they're in tier E and they have a fee of $5,760 a year. In that case, the fee is reasonably high. So this is how we would calculate based on the framework that I've described. That's how we would calculate a new set of fees and a new structure. The impact assessment - we, of course, need to look at what would be the impact of such a change on APNIC members themselves, on regular members, on NIRs and the members of NIRs. And we have to critically look at the impact on the APNIC budget, that's one of the drivers for this change. Cheaper revenues are possible to calculate on which we know about APNIC's membership and address holdings. Here is a table which shows the total current distribution of about 1,160 or so APNIC members showing how many members fit into each prefix here. For instance, at the /18 level, there are 89 members who have up to, above a /19 but up to and including a /18. Those 89 members would each pay $2,880 a year instead of the current $5,000. That's a reduction of $2,000 a year. The lines in blue represent members whose annual fee is reduced and the others are those whose annual fee has increased, all the way up to three ISPs in the very largest category, up to a /6. They've got more than a /7, this is three members who have at least two /8s in their address pool. And instead of paying $40,000 a year, this fee structure would have them paying $184,000 a year. We can clarify the idea here is that rather than NIRs paying as individual members, NIRs would pay according to the detailed address holdings of their own members. And we do, since we've been allocating address space directly from APNIC through NIRs to the members of NIRs, we do have at APNIC, a pretty good idea at least of the distribution of members of NIRs, at least of those who are active. In APJII's case, they have 33 members who would fit into tier three, two members in tier D and two members in tier E. There are a total of 37 members. And the table here shows the fee structure to those members who would pay of what I've just described. This doesn't take into account any discount for NIRs. So what the tables show in aggregate is there's 365 members in total that we know about of six NIRs. The total fee revenue from those would be $1.5 million without any discount. The overall impact on APNIC's budget - the discount that I'm modelling here in this example is a 50% discount for members of NIRs. The first line in this budget shows our current budget expectation for an NIR per-address fee revenue is $700,000. The example fee structure, with a 50% discount would give $778,000, which is an 11% increase. The normal regular APNIC members, we currently have a budget of $4.23 million for the year 2006. But the income from the example fee structure would be $4.83 million. The parameters that are selected in setting up this example were not selected randomly, they're selected to provide exactly that type of increase in the total and in the distribution of APNIC fees. But it's just an example and this is all driven by a spreadsheet and you can play with the numbers and try different scenarios and see what those scenarios would produce. So, that is the summary of an example fee structure of how we would assess it in terms of impact on members individually, impact on NIRs and members of NIRs, impact on APNIC's revenue but we have a few things to discuss. If we follow this type of framework, what should the parameters actually be to achieve what are the overall revenue expectations? Should the parameters be fixed or variable? APNIC has a fixed, stable fee structure for many years. One of the other NIRs, RIPE NCC, varies the fee structure each year according to the revenue requirements of their activity plan. So rather than being driving a budget on expected revenues, they drive the revenues from the expected budget and that would require us, if we took that approach, to take a variable approach to the parameters of the fee structure and vary them from each year or from time to time. The NIR discount, which is written in here, is 50%. It recognises the value of services provided by the NIR to members of the NIRs. So is that the appropriate discount level? Should the same discount apply to all NIRs? There are NIRs with not much experience and NIRs with a great deal of experience. So possibly the value of services provided by those NIRs should be assessed differently. It is a bit complicated but it would be possible to apply a scale of discounts to NIR fee structures. A question which came up in recent discussions about this is that NIRs have already paid fees for address space in expectations these are once and for all fees. Should there be a recognition of the fact this address space has been paid for already in a sense? And finally, as you saw, there's quite a significant impact from this type of fee structure and vision, potentially, on quite a few of the APNIC members. To introduce that in one single step would be quite dramatic a change and so in order to ease that transition, should it be, should a new structure be introduced immediately or possibly over one or more years to control the budget impact of that? So the next steps are that were decided at the time - a special session that we had on this during the week was to establish a mailing list which has been set up - wg-apnic-fees. Anyone can report back to this. It may not happen at APNIC but maybe later. That is where we're at with the fee structure discussion. So I'd be very interested in answering any questions or hearing any comments about this topic. CHRIS CHAUNDRY: Just a question, while things are, while things are being thrown up in the air, the other question that I have is as RIPE set its fee structure based on the euro, we still base our fee structure on the US dollar which is somewhat ironic, given it's a currency outside our region. Given that the majority of APNIC's expenses are incurred outside the country where it is resident now, would it not make sense to remove the variance of the exchange rates by actually charging in that particular currency? PAUL WILSON: OK. Thank you. TRENT O'CALLAGHAN: I work for a local ISP here in Perth. One side of this that you haven't considered is the network engineering side of pricing in that aggregation is forced on all those smaller ones by not getting discount at the moment. I'd like that to be considered as you go forward? PAUL WILSON: Aggregation of the routing level? TRENT O'CALLAGHAN: And exhausting of the address space is slowed by forcing people to go through larger blocks possible? Yeah. It's a point of view that I'd like to see considered. PAUL WILSON: OK. Well, that's - I'm not sure about the address space aggregation issue. There is an issue of business aggregation which is something that happens in this sort of acquisitions. It's another problem we have with the fee structure. If we have two members that are aggregated at a particular level, they may have been responsible for two sets of annual fees. A small minority of cases does the aggregated business have enough address space to bump them up to the next fee level? You have an economy of scale which is having a detrimental impact and an unpredictable impact as well in that sense. TRENT O'CALLAGHAN: If I simplify the point I'm trying to make, the size of the routing table has a size of the memory in routers impact. And so the smaller we keep the routing tables, the more aggregated, the cheaper we keep it, the cheaper the Internet is to engineer. PAUL WILSON: If there are no other comments, I suspect it might be coffee time. There's time for us all to think about these and chat about them over the coffee break, which is scheduled for 10:30 to 11:00. There should be time for open mic and question and answer sessions later. All of this is up for consideration and discussion. I would encourage you all, particularly the APNIC members in the audience here, to have a think and a chat about this and maybe some questions and issues will occur to you after that. So let's take a coffee break then and come back at 11. OK, thanks.