RE: [lir-wg] IPv6 assignments to RIPE itself
-----Original Message----- From: Alexander Gall [mailto:gall@switch.ch] Sent: 15 January 2003 09:15 To: Gert Doering Cc: Stephane Bortzmeyer; Jeroen Massar; 'Wilfried Woeber, UniVie/ACOnet'; lir-wg@ripe.net Subject: Re: [lir-wg] IPv6 assignments to RIPE itself
[Probably opening a can of worms here...]
Yuk, worms taste bad .... ;-) <snip>
I agree. The numerics in section 4 of RFC3177 assume that the top 45 bits in 2000::/3 can be utilized with an H ratio of 0.25 (giving on the order of 10^11 /48). IMHO, the problem with the current allocation policy is that it is a lot more conservative than RFC3177 suggests while still holding on to the /48-for-everybody rule. The relatively small LIR allocations create a level of scarcity in the number of /48's, which is enough to make people believe that giving a student as much address space as her entire University is just crazy. However, the whole point of RFC3177 was that this should be completely irrelevant.
It *is* completely irrelevant. Allocate /48s, exhaust existing RIR allocation, get more addresses from RIR. I don't see the problem. -- Mat.
On Wed, 15 Jan 2003 matthew.ford@bt.com wrote:
It *is* completely irrelevant.
Allocate /48s, exhaust existing RIR allocation, get more addresses from RIR.
I don't see the problem.
We're supposed to be conserving route announcements too... IP address space is not the only issue. aid -- Adrian Bool | http://noc.vianw.net/ Director, Global Core Network | tel://+44.1925.484061/ VIA NET.WORKS Inc. | noc://+49.203.3093.1111/
Hi, On Wed, Jan 15, 2003 at 11:25:12AM -0000, matthew.ford@bt.com wrote:
Allocate /48s, exhaust existing RIR allocation, get more addresses from RIR.
I don't see the problem.
The current way that the RIRs and IANA allocates space *is* a problem, because it leads to "multiple prefixes per LIR", which is bad bad bad bad. Alexander Gall has summarized it pretty well - if we want to give out /48s freely, then the quite conservative RIR->LIR allocation policy currently in place *hurts*. As for the argument "are universities ISPs"? Yes, at least over here, a fair number of them are providing IP connectivity to the student's hostels via leased line/ethernet, and to all other students via ISDN/Modem dialup. So for all address management purposes, they are ISPs. Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 55593 (55180) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
On Wed, Jan 15, 2003 at 13:56:18 +0100, Gert Doering wrote:
Alexander Gall has summarized it pretty well - if we want to give out /48s freely, then the quite conservative RIR->LIR allocation policy currently in place *hurts*.
Very true if you mean that you cannot build a reasonable hierarchy.
As for the argument "are universities ISPs"? Yes, at least over here, a fair number of them are providing IP connectivity to the student's hostels via leased line/ethernet, and to all other students via ISDN/Modem dialup. So for all address management purposes, they are ISPs.
This is true in the Netherlands too. Yes, I think those should be treated as ISPs, probably getting a prefix (>> /48) from their NRN. rvdp
Hi, On Wed, Jan 15, 2003 at 02:15:42PM +0100, Ronald van der Pol wrote:
On Wed, Jan 15, 2003 at 13:56:18 +0100, Gert Doering wrote:
Alexander Gall has summarized it pretty well - if we want to give out /48s freely, then the quite conservative RIR->LIR allocation policy currently in place *hurts*. Very true if you mean that you cannot build a reasonable hierarchy.
Yup. This is mostly the point behind my rantings. Trying to build a resonable hierarchy through some levels of regional aggregation inside my network, and then aggregation through 1-2 levels of resellers. The other point is that one of the main arguments in that RFC is "if a customer changes ISPs, they will always get the same size prefix (a /48)", which is just not working if customers can very liberally get more than a /48 to account for "another-level-down end sites". So we're back to the address space haggling days, just argueing about the number of /48s instead the number of single IPs. So I still think that the concept of "one /48 for each site" without a proper definition of "site" is flawed. And yes, it's arguably pretty impossible to give a working definiton.
As for the argument "are universities ISPs"? Yes, at least over here, a fair number of them are providing IP connectivity to the student's hostels via leased line/ethernet, and to all other students via ISDN/Modem dialup. So for all address management purposes, they are ISPs.
This is true in the Netherlands too. Yes, I think those should be treated as ISPs, probably getting a prefix (>> /48) from their NRN.
See above :-) Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 55593 (55180) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Hello! Gert Doering wrote: [...]
The other point is that one of the main arguments in that RFC is "if a customer changes ISPs, they will always get the same size prefix (a /48)", which is just not working if customers can very liberally get more than a /48 to account for "another-level-down end sites". So we're back to the address space haggling days, just argueing about the number of /48s instead the number of single IPs. [...]
A /48 isn't the same as an v4-IP. Using a /48, one has 65536 different /64-IPs which is really enough for most applications. Less people will request more than one /48 this way and it is and should be the job of the LIR to filter these who want as many IPs as possible just for fun. MfG / Regards, S.Willing ************************************************************************ M / W / S Sebastian Willing http://www.mops.net Technical director Telefon: 01803/684310578 e-Mail: s.willing@mops.net Telefax: 01805/445169111 SW88-RIPE ************************************************************************
Hi, On Wed, Jan 15, 2003 at 02:58:25PM +0100, Sebastian Willing wrote:
Gert Doering wrote: [...]
The other point is that one of the main arguments in that RFC is "if a customer changes ISPs, they will always get the same size prefix (a /48)", which is just not working if customers can very liberally get more than a /48 to account for "another-level-down end sites". So we're back to the address space haggling days, just argueing about the number of /48s instead the number of single IPs. [...]
A /48 isn't the same as an v4-IP.
I'm aware of that.
Using a /48, one has 65536 different /64-IPs which is really enough for most applications. Less people will request more than one /48 this way and it is and should be the job of the LIR to filter these who want as many IPs as possible just for fun.
The design of the policy is "give everybody a /48 without asking" so that there is no *need* for discussion, weeding out "unwarranted applications" and "if you don't give it to me, I go to $otherisp who will!". This contradicts the so-far result of the ongoing discussion, which is "if someone has some sort of semi-independent networks connected, they can get a /48 on their own, so the upstream has to get more space"... Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 55593 (55180) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
On Wed, Jan 15, 2003 at 14:49:37 +0100, Gert Doering wrote:
The other point is that one of the main arguments in that RFC is "if a customer changes ISPs, they will always get the same size prefix (a /48)", which is just not working if customers can very liberally get more than a /48 to account for "another-level-down end sites". So we're back to the address space haggling days, just argueing about the number of /48s instead the number of single IPs.
I don't agree. It's not just a customer. It's an ISP. If an ISP wants to switch from upstream provider, that's a big job. And some negotiation about prefix delegation is part of that. _End_ customers will get a /48. If they change ISP, they get a /48 again. Really big enterprise (end) customers with two or three /48s are not guaranteed to get the same amount of /48s from a new ISP. But I guess they will have a strong negotiating position.
So I still think that the concept of "one /48 for each site" without a proper definition of "site" is flawed. And yes, it's arguably pretty impossible to give a working definiton.
Yes, that's true. But "end customer" <--> "ISP" relations are pretty clear. Those will have /48 assignments. rvdp
Hi, On Wed, Jan 15, 2003 at 03:10:23PM +0100, Ronald van der Pol wrote:
On Wed, Jan 15, 2003 at 14:49:37 +0100, Gert Doering wrote:
The other point is that one of the main arguments in that RFC is "if a customer changes ISPs, they will always get the same size prefix (a /48)", which is just not working if customers can very liberally get more than a /48 to account for "another-level-down end sites". So we're back to the address space haggling days, just argueing about the number of /48s instead the number of single IPs.
I don't agree. It's not just a customer. It's an ISP. If an ISP wants to switch from upstream provider, that's a big job. And some negotiation about prefix delegation is part of that.
_End_ customers will get a /48. If they change ISP, they get a /48 again.
Really big enterprise (end) customers with two or three /48s are not guaranteed to get the same amount of /48s from a new ISP. But I guess they will have a strong negotiating position.
Just calling those parties ISPs will not solve the dilemma. A big company that wants to give IP connectivity to their employees (like the RIPE NCC does) is not an "ISP" in the classic sense, as it's not their main business. On the other hand, if you call everybody that happens to offer an ISDN S0 for dialup to their employees purposes an ISP, then most of our business customers could be called ISPs - which defeats the "one /48 for end sites" rule again. Of course there are some customers that make their money doing ISP business (that is: give paying customers who are different legal entities (!) access to the internet), and I have no problem with them getting a /36 or whatever. The problem is the class of customers like the RIPE NCC.
So I still think that the concept of "one /48 for each site" without a proper definition of "site" is flawed. And yes, it's arguably pretty impossible to give a working definiton.
Yes, that's true. But "end customer" <--> "ISP" relations are pretty clear. Those will have /48 assignments.
Well, for those "end customers" that are "end customers", assigning them a /48 is covered pretty clearly by the current policy. I still don't think that it's very easy to define a given business relation as "this is an end-end-end customer", and "that is an ISP". Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 55593 (55180) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
On Wed, Jan 15, 2003 at 15:37:27 +0100, Gert Doering wrote:
Just calling those parties ISPs will not solve the dilemma. A big company that wants to give IP connectivity to their employees (like the RIPE NCC does) is not an "ISP" in the classic sense, as it's not their main business.
On the other hand, if you call everybody that happens to offer an ISDN S0 for dialup to their employees purposes an ISP, then most of our business customers could be called ISPs - which defeats the "one /48 for end sites" rule again.
I agree. Maybe the difference is how the home network is connected? For most people, the normal conenctivity is via traditional ISPs. They get a /48 from that ISP. They may also have some kind of connectivity (VPN or dialup) to their company network. But I think that's mostly used for "secure" access to company services. Usually one /64 out of the company prefix will be enough. The company does not need to assign /48s to its employees. University xDSL, cable and fiber networks are different. Usually, it's the only connectivity for the students or employee. And they prefer to get /48. In this case the university can be seen as an ISP. I think some uncertainty comes from the fact that most ISPs are still not offering IPv6 services and people are trying to get an IPv6 prefix somewhere else.
I still don't think that it's very easy to define a given business relation as "this is an end-end-end customer", and "that is an ISP".
Thinking more about it, I agree. rvdp
The current way that the RIRs and IANA allocates space *is* a problem, because it leads to "multiple prefixes per LIR", which is bad bad bad bad.
I agree if the resulting address blocks can not be announced as a single aggregate. I was, however, under the impression that when a RIR makes an IPv6 allocation to a LIR, it leaves "room to grow" in the address space, so that some of the subsequent allocations can still be announced with a single routing announcement. This is different to the practices used in the IPv4 space, partly, I guess, because there is room to use that practice. Did this change and/or am I mis-remembering? Regards, - Håvard
Hi, On Wed, Jan 15, 2003 at 03:40:00PM +0100, Havard Eidnes wrote:
The current way that the RIRs and IANA allocates space *is* a problem, because it leads to "multiple prefixes per LIR", which is bad bad bad bad.
I agree if the resulting address blocks can not be announced as a single aggregate.
I was, however, under the impression that when a RIR makes an IPv6 allocation to a LIR, it leaves "room to grow" in the address space, so that some of the subsequent allocations can still be announced with a single routing announcement. This is different to the practices used in the IPv4 space, partly, I guess, because there is room to use that practice. Did this change and/or am I mis-remembering?
RIPE allocates the /32 out of a reserved /29. Which leaves quite some amount to grow, but I still think the whole distribution down from the top is still too much based on conservation-and-slow-start-thinking. <next rant> Why are we messing around with /23s being given from IANA to the RIRs in the first place? That way, filtering by region will quicky get REALLY messy, and I really can't see any useful reason to do so. A useful way would be to do /12s or so (IPv4 does /8s, since we're only operating in 001 the equivalent would be a /11, and nibble-aligning gives a /12). Or maybe a /16, which looks much nicer. </next rant> Gert Doering -- NetMaster -- Total number of prefixes smaller than registry allocations: 55593 (55180) SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
participants (6)
-
Adrian Bool -
Gert Doering -
Havard Eidnes -
matthew.ford＠bt.com -
Ronald van der Pol -
Sebastian Willing