Re: (ngtrans) Fw: [apnic-announce] IPv6 Policy Document Revision
Brian, At 10:29 -0500 7/7/00, Brian E Carpenter wrote: ...snip...
A /64 for dialup is also too rigid because the way technology is going a /64 is not going to be enough subnets for what wiill be a dial up connection with a large lan behind it.
Indeed. But that isn't an issue the RIRs need to think about.
It is not the RIRs trying to force variable length prefixes. At the RIPE meeting in Budapest and the ARIN meeting in Calgary we reported what the outcome of conversations with IETF people was (the /48, /56, /64 options for allocation). This seemed to be a reasonable way of doing it for the IPv6/ngtrans IETF people and to the RIR people present in Adelaide. At both the ARIN and RIPE meetings, ISPs (the people who will use the address space, after all) were the ones that suggested variable length prefixes for allocation and the RIRs must go by the community consensus (BTW, at the RIPE meeting no consensus was reached, with both the variable length and the 3 fixed lengths having supporters). A second issue is to think about a way of getting the IETF IPv6 people and the 3 RIR communities to talk to each other at the same, otherwise we are entering a loop, with at least 4 separate discussions and each dependant on the other 3. Joao
Brian
Of course we should talk, the next opportunity is the Pittsburgh IETF. However, let's be careful about deducing general policy from what today's ISPs believe, based on IPv4 experience. Things are truly different in IPv6. Brian Joao Luis Silva Damas wrote:
Brian,
At 10:29 -0500 7/7/00, Brian E Carpenter wrote: ...snip...
A /64 for dialup is also too rigid because the way technology is going a /64 is not going to be enough subnets for what wiill be a dial up connection with a large lan behind it.
Indeed. But that isn't an issue the RIRs need to think about.
It is not the RIRs trying to force variable length prefixes. At the RIPE meeting in Budapest and the ARIN meeting in Calgary we reported what the outcome of conversations with IETF people was (the /48, /56, /64 options for allocation). This seemed to be a reasonable way of doing it for the IPv6/ngtrans IETF people and to the RIR people present in Adelaide.
At both the ARIN and RIPE meetings, ISPs (the people who will use the address space, after all) were the ones that suggested variable length prefixes for allocation and the RIRs must go by the community consensus (BTW, at the RIPE meeting no consensus was reached, with both the variable length and the 3 fixed lengths having supporters).
A second issue is to think about a way of getting the IETF IPv6 people and the 3 RIR communities to talk to each other at the same, otherwise we are entering a loop, with at least 4 separate discussions and each dependant on the other 3.
Joao
Brian
At 06:59 PM 7/10/00, Brian E Carpenter wrote:
Of course we should talk, the next opportunity is the Pittsburgh IETF.
However, let's be careful about deducing general policy from what today's ISPs believe, based on IPv4 experience. Things are truly different in IPv6.
Brian
Brian, we - the RIRs - are in no position to tell the ISPs that they are stupid and that they "don't understand". I think even the mighty IAB should be a little careful before going that route; it is the ISPs, after all, who convert integers to address space by routing traffic based on it. Policies will change as we go along and all those involved understand more and get more comfortable with things. Daniel
Daniel Karrenberg wrote:
At 06:59 PM 7/10/00, Brian E Carpenter wrote:
Of course we should talk, the next opportunity is the Pittsburgh IETF.
However, let's be careful about deducing general policy from what today's ISPs believe, based on IPv4 experience. Things are truly different in IPv6.
Brian
Brian,
we - the RIRs - are in no position to tell the ISPs that they are stupid and that they "don't understand". I think even the mighty IAB should be a little careful before going that route; it is the ISPs, after all, who convert integers to address space by routing traffic based on it.
Policies will change as we go along and all those involved understand more and get more comfortable with things.
Indeed, my words that you quote were written to avoid insulting the ISP's intelligence and experience... but IPv6 is truly different. Brian
Indeed, my words that you quote were written to avoid insulting the ISP's intelligence and experience... but IPv6 is truly different.
yes. it's not deployed. and to get it deployed, as bernard just said well on the ietf list, we need to make it *significantly* more attractive than the status quo, like five times as attractive. one attraction could be address space. and ipv6 has a lot of it, though, having been through some decades in this game it is hard to believe that this year's effectively-infinite will not be next year's (or decade's) joke. so it would be nice if we could have an ipv6 address space allocation policy which allowed users to perceive the benefits of the larger address space. and, imiho, we should err on the side of generosity. but we should not forget that we have four more wonderful marvels of ipv6 to find if we want it to sell. randy
Joao Luis Silva Damas wrote:
Brian,
At 10:29 -0500 7/7/00, Brian E Carpenter wrote: ...snip...
A /64 for dialup is also too rigid because the way technology is going a /64 is not going to be enough subnets for what wiill be a dial up connection with a large lan behind it.
Indeed. But that isn't an issue the RIRs need to think about.
It is not the RIRs trying to force variable length prefixes. At the RIPE meeting in Budapest and the ARIN meeting in Calgary we reported what the outcome of conversations with IETF people was (the /48, /56, /64 options for allocation). This seemed to be a reasonable way of doing it for the IPv6/ngtrans IETF people and to the RIR people present in Adelaide.
At both the ARIN and RIPE meetings, ISPs (the people who will use the address space, after all) were the ones that suggested variable length prefixes for allocation and the RIRs must go by the community consensus (BTW, at the RIPE meeting no consensus was reached, with both the variable length and the 3 fixed lengths having supporters).
A second issue is to think about a way of getting the IETF IPv6 people and the 3 RIR communities to talk to each other at the same, otherwise we are entering a loop, with at least 4 separate discussions and each dependant on the other 3.
The only practical reason I can see for variable-length prefixes is for the ISP to charge more money for allocation of any prefix shorter than a /64. Guaranteed, the average ISP (no, I won't tar you all with the same brush) if permitted to allocate a /64 won't allocate anything shorter unless lubricated with many times the amount of cash. Perhaps it sounds cynical, but that's based on my experiences with ISPs. Under IPv4, I can see how a scarce resource falls under the supply/demand paradigm. Let's not falsely inflate the value of a /48 to the end-user by making longer prefixes available. Those longer prefixes will simply become the norm, which was not our intent, was it? D
From: Joao Luis Silva Damas <joao@ripe.net> Subject: Re: (ngtrans) Fw: [apnic-announce] IPv6 Policy Document Revision
A second issue is to think about a way of getting the IETF IPv6 people and the 3 RIR communities to talk to each other at the same, otherwise we are entering a loop, with at least 4 separate discussions and each dependant on the other 3.
Please make the way. --Kazu
participants (6)
-
Brian E Carpenter -
dancer@zeor.simegen.com -
Daniel Karrenberg -
Joao Luis Silva Damas -
Kazu Yamamoto -
Randy Bush