- ipv6-wg - mailman.ripe.net

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by CJ Wittbrodt 08 Feb '02

08 Feb '02

Hi. Yes. But then they need an upstream provider that is willing to announce a single-homed customer /32 - which might just be much more expensive than giving the end site a /48 of their space. Which I could imagine as a way to let market regulate whether someone "needs" their own /32. Sure if ISPs are willing to filter them. If there are lots of them and they aren't aggregatable then the filters will be long. I suspect they'll end up being the swamp and everyone will end up carrying them even if they don't want to. History repeats itself on many levels... I am truly convinced that we can NOT solve the "I want to be independent" problem at registry level (or at least "not without doing real harm to some of the people we want to be 'in'"). Adding a rule like "you get a /32 only if you're multihomed" will make things worse, because "they" will then get a /32 *and* an AS number. I agree. But we're not currently saying that you get a /32 if you're multihomed, we're saying that you get a /32 if you're willing to pay money. How about you get a /32 if you can justify at least some reasonable level of utilization? That could be that you currently have a minimum allocation of IPv4 space from a registry already. It could be any number of things probably less restrictive than what was on the table until recently. Then sure you can get one of you are whoever, but at least you have some real justification for having a block from a registry and not your upstream. Maybe we should look at the people you refer to as wanting to be "in" and figure out what distinguishes them from just some end user who should have a /48 from their upstream...? ---CJ

3 3

How to reduce the junk applications?
by turchanyi.geza＠ln.matav.hu 07 Feb '02

07 Feb '02

Bob, the question here is the following: How to avoid that a boom of (virtual) LIRs destroys the whole Internet Registry structure? My proposal is simple: only such a LIR should get IPv6 addresses in 2002 and 2003 that already had experiences in IPv4 registration for at least two years. This might sound very restrictive, however it is better to start something new with people already having some experiences than creating a situation that we can not handle later on. In two years we can have lift this constraint. Best regards, Geza Turchanyi Bob Hinden <hinden(a)IPRG.nokia.com> dátum: 2002.02.06 20:20:49 Címzett: Gert Doering <gert(a)Space.Net> Másolat: "Alec H. Peterson" <ahp(a)hilander.com>, Brian E Carpenter <brian(a)hursley.ibm.com>, Thomas Narten <narten(a)us.ibm.com>, Stuart Prevost <stuart.prevost(a)btinternet.com>, global-v6(a)lists.apnic.net, ipv6-wg(a)ripe.net (Vakmásolat: Turchányi Géza/PKI/HTC2) Tárgy: Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy Gert, >In that 1/8th, we have 500 million /32s. I proposed to give out 2000 of >them per region. So roughly this is 0.001 per cent of the available >/32s in the 1/8ths that we're currently using. > >Yes, I vote for "let's waste that 0.001 per cent", and then reconsider. I think this is a good approach. To compare with the IPv4 history, many of the early allocations in IPv4 were the old Class A prefixes. Each Class-A prefix consumed 0.391% of the total IPv4 address space. To compare with the new IPv6 registry proposal, each IPv6 /32 prefix consumes 0.000000186% of the total IPv6 address space. This is roughly equivalent to the amount of address space used by a /29 prefix allocation in the IPv4 space. Bob - - This list (global-v6) is handled by majordomo(a)lists.apnic.net

1 0

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by CJ Wittbrodt 07 Feb '02

07 Feb '02

Hi, On Wed, Feb 06, 2002 at 02:25:46PM -0800, CJ Wittbrodt wrote: > It is a figure that is "large enough that nearly every active LIR today > can get an IPv6 allocation NOW" (having somewhat over 3000 LIRs in RIPE > land, of which some are not active any more, others have merged, and so > on), while at the other hand being small enough so that *if* this > turns out to be a mistake, it means "6000 'IPv6 swamp' prefixes in the > global routing table", and this is something the routers can handle. > > I just want to be clear. My understanding of what happened in the > meeting wasn't that each existing LIR could get one, but anyone who > became an LIR could get one. This means that existing LIRs can > indeed get one, but so can anyone who is willing to become an LIR. > That was my understanding. Yes, this is what was agreed upon. Every LIR, no matter whether old or new, could get a /32 by asking for it. I do not think that this will lead to a "land rush" on /32s (to answer that concern en passant). Why should it? What's the benefit for an end site? It gives them a provider block that won't be filtered. If they're multihomed this is a benefit. Most of the companies that want to be "independent" will find one way or the other to achieve this - either by announcing /48s all over the region and possibly the world, or by opening a LIR, or by claiming they want to be soooo multihomed (and maybe setting up a peering with some other "independent" company to prove it). We have to be able to solve *this* - teach 'em that BGP multihoming with "PI" space is just one of many solutions, and develop more attractive solutions - instead of hindering IPv6 progress any further. Why would they justify or claim anything? They can just become an LIR with no justification and get a /32. ---CJ

1 0

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by CJ Wittbrodt 07 Feb '02

07 Feb '02

Since the events of the recent RIPE working group meeting have not been discussed at either of the other regional policy forums, proposing something here may be somewhat premature. I do not believe, although things never cease to amaze me, that this will reach any sort of consensus within the ARIN region. I am not sure about the APNIC region. Based on some meetings with the European Government Advisory Council (just after the RIPE meeting) it is clear that it is important that we have a global policy. Is there maybe some way that we could come up with a compromise that would reach consensus in all three policy forums? Something other than requiring no justification for a /32? Thanks ---CJ Wittbrodt (ARIN Advisory Council and ASO Address Council member) From: Gert Doering <gert(a)Space.Net> Subject: Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment >>Global Policy Hi, from your comments, I gather you have not been to the RIPE IPv6/LIR policy meeting. So let's add a few comments (while waiting for James Aldridge to publish the "official" word on it): On Fri, Feb 01, 2002 at 02:50:38PM -0500, Thomas Narten wrote: > > 5.2.1. Initial allocation criteria [..] > The goal was to give a site a /32 if it can justify it will use > it. The word "immediately" (as in demonstrate an immediate need) is [..] > So the real issue here (and this comes up again in later parts of the > document) is what is a reasonable way to objectively evaluate a > request for address space that requires some guessing as to whether a > proposed plan will actually be carried out. If the time frame is too > long, it becomes easy to make optimistic plans that won't pan out, and > then the RIRs get into a different problem. Consensus on the IPv6/LIR policy meeting was "drop the criteria". To be precise, I proposed the following: - any LIR that is established (has done all the paperwork, paid their fees, and whatnot) and can document the need for one IPv6 address can get a /32. No further justification required. - to avoid a horrible mistake, every region is permitted to allow only assigment of 2000 /32s per region. So the maximum wastage is 6000 /32s (out of 500 million /32s in the 1/8th of the space we're talking about), and 6000 additional routes. After that, we're going to reconsider policy. There was concern from the other regional registries (ARIN and APNIC), but broad consensus from the people from the RIPE region. Reasoning (shortened): - why are we putting criteria there? To keep out "some that we do not want". Conservation is not an issue. Routing table growth might be influenced by this, or might be not, we don't know. - do we want major national research networks connecting something like "50 universities"? YES - will this research network meet any criteria based on "you must use up a big number of /48s, otherwise you can't get a /32"? NO, if you assign a /48 per university (which would be plenty!), because that means "you can only demonstrate a need for 50 /48s" On the other hand, if you say "I connect lots of private customers over DSL lines, using fixed IP addresses, giving each user a /48 (which is OK according to IETF guidelines)", reaching over 50 /48s is very easy. Does this mean the second example is "more worthy" to get a /32? Does it mean they will make "better use" of it? So all technical criteria based on /48 usage must fail, and criteria based on single IP usage will fail as well (due to the /48 rule). If we can't propose criteria that work, drop them - BUT limit the amount of damage that can be done. [..] > Or is it the 776 end site figure (i.e., too high)? Think of the research networks. One /48 per university would be "according to the /48 rule: each SITE gets a /48". Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 71770 (72395) SpaceNet AG Mail: netmaster(a)Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299 - - This list (global-v6) is handled by majordomo(a)lists.apnic.net

4 3

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by Thomas Narten 07 Feb '02

07 Feb '02

Gert Doering <gert(a)Space.Net> writes: > As for the routing table growth, yes, this is a very tricky problem, > but it cannot be solved at RIR level - the RIRs can not decide who is > an "important ISP" and who is a "less important ISP that should > prevented from having their own networks" (by being forced to use > upstream addresses from "important ISP" - renumbering is fairly easy > for end sites, but a PITA for an ISP plus all his customers). The RIRs may not be able to solve this problem, but they can surely adopt policies that will make it harder if not impossible to solve it. It is this aspect of the apparent proposal in RIPE that I have a BIG problem with. It is far from clear that we have to throw the baby out with the bath water. If folks have issues with the specific policies in the global document, then can we first see if we can fix them in a way that makes sense to all, and doesn't make addressing the routing table growth problem even harder than it already is? Please? Thomas

1 0

RE: [GLOBAL-V6] New draft available: IPv6 Address Allocation andAssignment Global Policy
by Ext-Pauli.Wihuri＠nokia.com 07 Feb '02

07 Feb '02

It has been very interesting to follow your discussion here. Thanks. I know my comments/question may go off topic, but... >On 7 Feb 2002, Robert Kiessling wrote: >>Sorry to ask since I wasn't at the meeting: >>What was said about larger allocation sizes >>than /32, e.g. for multinational ISPs for >>which a /32 does not provide sufficient >>space for reasonable aggregation? On 7 Feb 2002, Tim Chown wrote: >Or for a national provider who may want to >give a static /48 to a few hundred thousand >customers. How much of the address assignment >maths includes static vs dynamic provisioning? Or for a global multinational company (not being ISP/LIR/...) having around the globe offices and sites and multiple public/private network transfer points. Here are the alternatives thought of: a) Introduce one global big enough prefix with hierarchy classes like region/country/city/site/business_unit/function/network_type and reserve space enough to each class AND also requiring optimized routing solution with many! private/public network transfer points. (preferred I guess) b) Applying for different prefix from all regions and introduce geographical address allcoation scheme with those different prefixes from all various regions... (feasible and likely, but not liked) c) Applying for different prefixes from all countries and introduce geographical address allocation scheme with thos hundreds of different prefixes... (no thanks :-) d) Other, what? -- Pauli Wihuri,Nokia,IPv6 network program manager,(external from Hewlett-Packard)

1 0

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by CJ Wittbrodt 07 Feb '02

07 Feb '02

Gert Doering <gert(a)space.net> writes: > Yes. But then they need an upstream provider that is willing to announce > a single-homed customer /32 - which might just be much more expensive than > giving the end site a /48 of their space. Which I could imagine as a > way to let market regulate whether someone "needs" their own /32. Why? If it there was a significat effect of costs on prefix announcement, we wouldn't see the huge number of (pseudo-) PI prefixes in today's table. I don't see a technical difference between announcement of an IPv4 PI /24 and an IPv6 /32 which would imply a significant difference in costs. Thus you will get an upstream for your v6 /32 as easily as now do for your v4 /24. The proposal poses the interesting question of other conditions to becoming LIR apart from costs. Currently RIPE requires a minimum of imediately assigned address space. What if anyone wants to become LIR for IPv6 only? Was this discussed? Yes. If they want to come and get IPv6 only there is no minimum criteria. Only that you become an LIR. Others can correct me if I am wrong, but I don't think so. Mirjam specifically talked about how they have a minimum requirement for IPv4 space, but in the beginning they didn't. The requirement now is because folks figured out the "I can get one and not be an ISP" thing. So if you pay your money and become a LIR you get a /32. That's what was proposed. I am sure there are minutes of the meeting by now. Nope, just looked they're not there yet. > "For a limited period (the first 2000 requests in each RIR > region was suggested) > allocate a /32 to ANY LIR requesting their IPv6 allocation > (subject to documentation of first IPv6 assignment) This is what was presented in the working group summary. I wrote it down verbatim off of the overheads. There was a promise of the documentation of the whole discussion. I am sure that's taking a while since it was quite lengthy. Sorry to ask since I wasn't at the meeting: What was said about larger allocation sizes than /32, e.g. for multinational ISPs for which a /32 does not provide sufficient space for reasonable aggregation? That was touched upon, but not really talked about in detail. Anne from APNIC brought it up as one of her concerns. Don't worry about not being at the meeting. The weather was cold and wet. :-) ---CJ

1 0

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by CJ Wittbrodt 06 Feb '02

06 Feb '02

From: Gert Doering <gert(a)Space.Net> Subject: Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy Hi, On Wed, Feb 06, 2002 at 05:17:41PM +0000, Carlos Friacas wrote: > The following are some of my personal views about it... > > I agree with almost all of Gert's views on this, my only doubt goes to > the "2000 figure". Isnt it a bit small? Why not look at the actual routing > table, and project things a bit further? Why not multiply it by 5 or 6? The figure was not chosen in any highly scientific way. It is a figure that is "large enough that nearly every active LIR today can get an IPv6 allocation NOW" (having somewhat over 3000 LIRs in RIPE land, of which some are not active any more, others have merged, and so on), while at the other hand being small enough so that *if* this turns out to be a mistake, it means "6000 'IPv6 swamp' prefixes in the global routing table", and this is something the routers can handle. I just want to be clear. My understanding of what happened in the meeting wasn't that each existing LIR could get one, but anyone who became an LIR could get one. This means that existing LIRs can indeed get one, but so can anyone who is willing to become an LIR. That was my understanding. ---CJ

1 0

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by CJ Wittbrodt 06 Feb '02

06 Feb '02

Thomas, Comments are below: From: Thomas Narten <narten(a)us.ibm.com> Subject: Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment >>Global Policy I wasn't at the RIPE meeting, but I do have at least one big concern based on what I heard. > To be precise, I proposed the following: > - any LIR that is established (has done all the paperwork, paid their > fees, and whatnot) and can document the need for one IPv6 address > can get a /32. No further justification required. Does this mean end sites, those that are not transit providers become eligible for a /32? In other words, what is to prevent a land grab on what will largely be perceived as PI IPv6 address blocks? (I heard from someone who was at the meeting that the answer was a clear yes, end sites will be eligible.) I think we would be sending entirely the wrong message if end sites are able to get /32s. I was at the RIPE meeting. The consensus was that anyone willing to become an LIR and pay the fees could get a /32. Certainly within RIPE there are requirements for becoming an LIR, but I don't believe those requirements distinguish between being and end site or whatever. We certainly discussed this at the meeting. There is nothing to prevent an end user from becoming an LIR and getting a /32. Note also that my main concern has much less to do with utilization or how much address space we have (we do have a lot). It has much more to do with having registry policies that are supportive of aggregation of routing information in the long term. This will only happen if most end sites get address space from their providers. Allowing end sites to get big allocations now, will cause nothing but problems later. It risks replaying some of the bad history of IPv4. That's right. > - to avoid a horrible mistake, every region is permitted to allow only > assigment of 2000 /32s per region. So the maximum wastage is 6000 > /32s (out of 500 million /32s in the 1/8th of the space we're talking > about), and 6000 additional routes. > After that, we're going to reconsider policy. And once we have 6000 such /32s allocated, then what? We replay the haves vs. the have nots? Won't there be a lot of pressure then to continue giving out /32s freely, since current router technology supports more than 6K routes, etc.? What makes you think we'll be able to put in more conservative policies at this point? This is definitely an issue. One that folks at the meeting didn't want to hear > - why are we putting criteria there? To keep out "some that we do not > want". Conservation is not an issue. Routing table growth might > be influenced by this, or might be not, we don't know. Having RIR policies that make it harder to have good aggregatability of routes should not be done rashly. It is far easier to take steps now that limit potential problems, than try to reverse poor decisions after the fact. Yup. > - do we want major national research networks connecting something like > "50 universities"? YES So, can we come up with alternate criteria that support giving /32s to such entities that doesn't also allow just anyone to get a /32? This desire seems reasonable to me. But the proposed solution seems to gone way beyond what was needed, and has some real down sides. That was attempted to be proposed. I'd like to see this group come up with something more than just pay the fee and get a block. A portable no aggregatable block. Further, the way blocks are being handed out now they aren't geographically aggregatable either. So each registry just peels off the next block. There aren't blocks per registry. > Who are the ones that we want to keep out anyway? Why? To achieve what? End sites. If every end site gets its own /32, we will have another routing disaster. So much for IPv6 learning from the IPv4 experience. Well the argument is that they'll never come back for more. This means (believe me this isn't my argument) that the routing table should be smaller because everyone will only have one prefix. There are only as many multi-homed sites as there are ASNs, right? Anyway, this is an argument that I have heard proposed. Please remember it isn't my argument. ----CJ

1 0

Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
by Thomas Narten 06 Feb '02

06 Feb '02

I wasn't at the RIPE meeting, but I do have at least one big concern based on what I heard. > To be precise, I proposed the following: > - any LIR that is established (has done all the paperwork, paid their > fees, and whatnot) and can document the need for one IPv6 address > can get a /32. No further justification required. Does this mean end sites, those that are not transit providers become eligible for a /32? In other words, what is to prevent a land grab on what will largely be perceived as PI IPv6 address blocks? (I heard from someone who was at the meeting that the answer was a clear yes, end sites will be eligible.) I think we would be sending entirely the wrong message if end sites are able to get /32s. Note also that my main concern has much less to do with utilization or how much address space we have (we do have a lot). It has much more to do with having registry policies that are supportive of aggregation of routing information in the long term. This will only happen if most end sites get address space from their providers. Allowing end sites to get big allocations now, will cause nothing but problems later. It risks replaying some of the bad history of IPv4. > - to avoid a horrible mistake, every region is permitted to allow only > assigment of 2000 /32s per region. So the maximum wastage is 6000 > /32s (out of 500 million /32s in the 1/8th of the space we're talking > about), and 6000 additional routes. > After that, we're going to reconsider policy. And once we have 6000 such /32s allocated, then what? We replay the haves vs. the have nots? Won't there be a lot of pressure then to continue giving out /32s freely, since current router technology supports more than 6K routes, etc.? What makes you think we'll be able to put in more conservative policies at this point? > - why are we putting criteria there? To keep out "some that we do not > want". Conservation is not an issue. Routing table growth might > be influenced by this, or might be not, we don't know. Having RIR policies that make it harder to have good aggregatability of routes should not be done rashly. It is far easier to take steps now that limit potential problems, than try to reverse poor decisions after the fact. > - do we want major national research networks connecting something like > "50 universities"? YES So, can we come up with alternate criteria that support giving /32s to such entities that doesn't also allow just anyone to get a /32? This desire seems reasonable to me. But the proposed solution seems to gone way beyond what was needed, and has some real down sides. > Who are the ones that we want to keep out anyway? Why? To achieve what? End sites. If every end site gets its own /32, we will have another routing disaster. So much for IPv6 learning from the IPv4 experience. Thomas

1 0