Re: [GLOBAL-V6] New draft available: IPv6 Address Allocation and Assignment Global Policy
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@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@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@lists.apnic.net
Just a note that the next APNIC Policy meeting is not too far away. It is being held in conjunction with APRICOT in Bangkok, Thailand. The new v6 policy draft (recently posted to this list) as well as the recent consensus from the RIPE community will be discussed in the Address Policy SIG on Wednesday 6th March. Full details of the programme can be found at: http://www.apnic.net/meetings/13/programme/index.html cheers, Anne _____________________________________________________________________ On Tue, 5 Feb 2002, CJ Wittbrodt wrote:
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@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@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@lists.apnic.net - - This list (global-v6) is handled by majordomo@lists.apnic.net * Mailing List: hostmaster-staff * * Handled by majordomo@staff.apnic.net *
Hi, On Tue, Feb 05, 2002 at 02:54:57PM -0800, CJ Wittbrodt wrote:
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.
Uh, I did not mean to "propose anything here". I was just reporting what we discussed at the RIPE meeting, and the consensus that was reached there. As the idea of conservation and very restrictively handling everything is rooted very deeply in all the registry people, I tried to explain *why* this new approach was (and is, I believe) seen as a good thing.
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?
The RIPE forum discussed alternative ways to justify the /32, and did not come up with anything that would "keep out the bad, but permit the good". Who are the ones that we want to keep out anyway? Why? To achieve what? Please take your time to consider these questions, and then try to find a poliy that will let in only "the right ones". After that, consider what harm could be done by being less restrictive than what we are used to. Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 71770 (72395) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Cathy, I'm still quite surprised by the way in which the fully justified conservatism of the registries for IPv4 space is being over-extrapolated to IPv6 space. If you look objectively at the argument that Gert gives, and consider how the size of the IPv6 prefix space compares to the total IPv4 space, /32 just isn't risky, and it gets rid of yet another judgement call. (I wasn't at the RIPE meeting either, but I did hear Mirjam talk on this topic yesterday.) Brian CJ Wittbrodt wrote:
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@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@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
--On Wednesday, February 06, 2002 10:13 +0100 Brian E Carpenter <brian@hursley.ibm.com> wrote:
Cathy,
I'm still quite surprised by the way in which the fully justified conservatism of the registries for IPv4 space is being over-extrapolated to IPv6 space. If you look objectively at the argument that Gert gives, and consider how the size of the IPv6 prefix space compares to the total IPv4 space, /32 just isn't risky, and it gets rid of yet another judgement call.
The issue that I and many other people on the ARIN AC continue to come back to is that this sounds a lot like the logic that was used when deciding how to originally allocate IPv4 address space. The fact that people have figured out how to conserve IPv4 address space (dynamic addressing, HTTP/1.1 Host: headers, etc) is a tribute to the fact that it _can_ be done if need be. But that's just the point, it was only done out of _necessity_, not because it was the right thing to do. The concern is that if we go back to an attitude of 'we will never run out of addresses' then we will find ourselves in a very unfortunate situation once again. Memory can be very short. We must learn from history, or we will be doomed to repeat it. Alec -- Alec H. Peterson -- ahp@hilander.com Chief Technology Officer Catbird Networks, http://www.catbird.com
Hi, On Wed, Feb 06, 2002 at 09:38:46AM -0700, Alec H. Peterson wrote:
I'm still quite surprised by the way in which the fully justified conservatism of the registries for IPv4 space is being over-extrapolated to IPv6 space. If you look objectively at the argument that Gert gives, and consider how the size of the IPv6 prefix space compares to the total IPv4 space, /32 just isn't risky, and it gets rid of yet another judgement call.
The issue that I and many other people on the ARIN AC continue to come back to is that this sounds a lot like the logic that was used when deciding how to originally allocate IPv4 address space.
I have heard that argument before, and my response boils down to "how much harm can we do?". We are currently allocating 1/8th of the IPv6 space. So if we *really* get everything wrong, we can try 7 additional times. 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. If we give every single LIR in the world a /32, long before the /32s run out, the *handling* of those registries will make be a HUGE problem (imagine a RIR having to handle 2^20 = 1 million!) LIRs. I do not opt for carelessness (I'd like a /8 for me, of course), but the numbers are so big in comparison to the number of potential candidates, and the scaling issues that would hit other places long before the IPv6 space runs out, that this really shouldn't be an issue. [..]
Memory can be very short. We must learn from history, or we will be doomed to repeat it.
We have enough addresses this time, so the lesson HAS been learned (yes, I know, everybody will shout at me about "640kbyte is enough", but there's only so much addresses that you can put on every square inch). (If we want to be conservative, stop putting /64s on local broadcast media, and stop handing out /48s to anybody who happens to have two different subnets in his home network. THAT is "wastive" - but everybody agrees that "we have enough /48s". Be consequent, and check the above numbers). Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 71770 (72395) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Hello. 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? Gert's math refers about 0.001 of 1/8th... 0.005 or even 0.010 seems to me perfectly reasonable... I also have one question and probably its because i havent been at RIPE-41 or because i didnt pay attention to some document... i now hear about /32s... People that already has a /35 will have their allocation somehow morphed in a /32, or the /32 will be just applied onward? Perhaps its a senseless question, but will probably have a short & fast answer... ;-) Thanks, ./Carlos "Networking is fun!" ------------------- <cfriacas@fccn.pt>, CMF8-RIPE, Wide Area Network WorkGroup http://www.fccn.pt F.C.C.N. - Fundacao para a Computacao Cientifica Nacional fax: +351 218472167 On Wed, 6 Feb 2002, Gert Doering wrote:
Hi,
On Wed, Feb 06, 2002 at 09:38:46AM -0700, Alec H. Peterson wrote:
I'm still quite surprised by the way in which the fully justified conservatism of the registries for IPv4 space is being over-extrapolated to IPv6 space. If you look objectively at the argument that Gert gives, and consider how the size of the IPv6 prefix space compares to the total IPv4 space, /32 just isn't risky, and it gets rid of yet another judgement call.
The issue that I and many other people on the ARIN AC continue to come back to is that this sounds a lot like the logic that was used when deciding how to originally allocate IPv4 address space.
I have heard that argument before, and my response boils down to "how much harm can we do?".
We are currently allocating 1/8th of the IPv6 space. So if we *really* get everything wrong, we can try 7 additional times.
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.
If we give every single LIR in the world a /32, long before the /32s run out, the *handling* of those registries will make be a HUGE problem (imagine a RIR having to handle 2^20 = 1 million!) LIRs.
I do not opt for carelessness (I'd like a /8 for me, of course), but the numbers are so big in comparison to the number of potential candidates, and the scaling issues that would hit other places long before the IPv6 space runs out, that this really shouldn't be an issue.
[..]
Memory can be very short. We must learn from history, or we will be doomed to repeat it.
We have enough addresses this time, so the lesson HAS been learned (yes, I know, everybody will shout at me about "640kbyte is enough", but there's only so much addresses that you can put on every square inch).
(If we want to be conservative, stop putting /64s on local broadcast media, and stop handing out /48s to anybody who happens to have two different subnets in his home network. THAT is "wastive" - but everybody agrees that "we have enough /48s". Be consequent, and check the above numbers).
Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 71770 (72395)
SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
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.
Gert's math refers about 0.001 of 1/8th... 0.005 or even 0.010 seems to me perfectly reasonable...
Yes, but that would mean 60.000 "bad routes" (if it turns out to be wrong), that's where the 3x2000 number came from.
I also have one question and probably its because i havent been at RIPE-41 or because i didnt pay attention to some document... i now hear about /32s... People that already has a /35 will have their allocation somehow morphed in a /32, or the /32 will be just applied onward? Perhaps its a senseless question, but will probably have a short & fast answer... ;-)
It's in the draft somewhere: existing /35s get extended to /32s, no questions asked besides "do you want this?". As far as I can remember, there was consensus on this point in the RIPE group. Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 71770 (72395) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Gert Doering writes:
We are currently allocating 1/8th of the IPv6 space. So if we *really* get everything wrong, we can try 7 additional times.
This argument always leaves me wondering what you plan to do with the existing routing table. Dump it on some sort of flag day? Carry twice the number of routes during a transition period? Not really workable if you expect table growth to be the limiting factor, no? Regards, -- Niels.
Hi, On Thu, Feb 07, 2002 at 03:06:07PM +0100, Niels Bakker wrote:
Gert Doering writes:
We are currently allocating 1/8th of the IPv6 space. So if we *really* get everything wrong, we can try 7 additional times.
This argument always leaves me wondering what you plan to do with the existing routing table. Dump it on some sort of flag day? Carry twice the number of routes during a transition period? Not really workable if you expect table growth to be the limiting factor, no?
I don't know what the limiting factor will be. I'm mainly argueing that *conservation* is really a big non-issue here, but *if* we are wrong in assuming this, we are able to try again. 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). Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 71770 (72395) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
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" -
If the RIR have not to check small allocations anymore, they can chack if an ISP is big or not. Number of transit, peering, bandwidth, number of customer... define if an ISP is big or not, and then RIR can decide if it *need* to be independent An ISP with a single 34Mbps line, a single transit and no peering can renumber.
renumbering is fairly easy for end sites, but a PITA for an ISP plus all his customers).
When an ISP is small, renumbering is not a really big problem. DNS changes can be automated. Router configuration, if it was good made are not a problem either. Host configuration is not a problem at all in v6. -- Xavier Henner Responsable de l'exp�rimentation IPv6 Nerim -- Fournisseur d'acc�s � Internet URL: <http://www.nerim.net/>
On Wed, Feb 06, 2002 at 09:38:46AM -0700, Alec H. Peterson wrote: Alec, all,
The issue that I and many other people on the ARIN AC continue to come back to is that this sounds a lot like the logic that was used when deciding how to originally allocate IPv4 address space.
Nobody wants to be the "640k is enough for anybody" advocate, and you're quite right to suggest that if we get this wrong, there will be unfortunate consequences for everyone. However, there is clearly a tension between the urgent requirement that some people have /today/, /now/, for address space, and the policy convergence of checks and balances that the RIRs would like to see. With all due respect, and I have a lot of respect for the RIR people, we have been talking about IPv6 policies for years, and working together across continents is difficult. Surely we should not be concentrating so ferociously on not getting it wrong, and instead concentrate on refining towards the right? Niall -- Enigma Consulting Limited: Security, UNIX and telecommunications consultants. Address: Floor 2, 45 Dawson Street, Dublin 2, Ireland. http://www.enigma.ie/
participants (4)
-
Alec H. Peterson -
Anne Lord -
Brian E Carpenter -
Carlos Friacas -
CJ Wittbrodt -
Gert Doering -
Niall Richard Murphy -
Niels Bakker -
Xavier Henner