[pekkas@netcore.fi: RIR ISP to end-user address allocation policy?]
Here's a message from the IPv6 WG listserv. This should start some discussion.. John -- John WELLS INRIA Rhône-Alpes équipe Planète - ENSIMAG 3A/Télécomm et Réseaux Virginia Tech MS/Computer Engineering - Networking and Visualization Lab Clé public: finger wells@mansoun.inrialpes.fr <interne>
Hi, On Wed, May 16, 2001 at 02:52:07PM +0200, John Wells wrote:
Here's a message from the IPv6 WG listserv. This should start some discussion..
Why? [..]
Hello all,
In a very recent RIPE meeting 1st May, Mirjam K|hne and Randy Bush presented the following on on IPv6 Address allocation policies:
http://www.ripe.net/ripe/meetings/archive/ripe-39/presentations/ipv6develop/
Among others, on slide 8, "ISP to Customer" there is:
--- * IAB/IESG recommended /48. * Use a /128 where it is absolutely known that one and only one device is required, e.g. dialup [<--!!!!!!!] * Use a /64 when sure net will not be subnetted, e.g. a mobile phone given 802.11, bluetooth, etc. ---
I find this thinking, or at least the examples very flawed.
Anyone want to start implementing NATv6 for people whose ISP refuses to give enough addresses to you can't (sub)network your home?
The wording is perfectly clear: if you have more than one device, the ISP MUST give you a /64 (under that policy). If you have more than one subnet, the ISP MUST give you a /48. [..] Gert Doering -- NetMaster -- SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
On Wed, 16 May 2001, Gert Doering wrote: [snip]
--- * IAB/IESG recommended /48. * Use a /128 where it is absolutely known that one and only one device is required, e.g. dialup [<--!!!!!!!] * Use a /64 when sure net will not be subnetted, e.g. a mobile phone given 802.11, bluetooth, etc. ---
I find this thinking, or at least the examples very flawed.
Anyone want to start implementing NATv6 for people whose ISP refuses to give enough addresses to you can't (sub)network your home?
The wording is perfectly clear: if you have more than one device, the ISP MUST give you a /64 (under that policy). If you have more than one subnet, the ISP MUST give you a /48.
Yes. I can already see the pricing: One device: 20$ 2+ devices: 40$ Network of devices: 60$ Would this kind of "voluntary" assignment work in practise? I wouldn't bet a penny on it; ISP's would just do /128 and write the customer contracts so that the other methods would not be possible in practise. Also, I do not see how a wireless device needs /64 when dial-up wouldn't. I think it's common today that e.g. 2 home computers share an internet connection. With mobiles and other gadgets going IP this might increase. If /128 assignment is the default, this by default would limit the options people are given. Someone frustrated with different pricing would write a hack to do NATv6. I don't think this should be encouraged. The rules should be set so that the default assignment is at least /64. -- Pekka Savola "Tell me of difficulties surmounted, Netcore Oy not those you stumble over and fall" Systems. Networks. Security. -- Robert Jordan: A Crown of Swords
I agree, giving customers a /128 almost guarantees that customers won't be able to run a private network. NATs are difficult to set up on IPv4 (don't ask your mom to do it), but on top of that they're not particularly well liked on IPv6. This may be somewhat of a newbie question, but would all v6 networks want to have DNS delegation, including PANs? I could see an ISP giving everyone a /64, but only delegating DNS authority to those /64s who register themselves as networks. I guess it depends on how DNS will pan out in IPv6. Regards, John mercredi, le 16 mai 2001 à 16h11 +0300, Pekka Savola a écrit :
On Wed, 16 May 2001, Gert Doering wrote: [snip]
--- * IAB/IESG recommended /48. * Use a /128 where it is absolutely known that one and only one device is required, e.g. dialup [<--!!!!!!!] * Use a /64 when sure net will not be subnetted, e.g. a mobile phone given 802.11, bluetooth, etc. ---
I find this thinking, or at least the examples very flawed.
Anyone want to start implementing NATv6 for people whose ISP refuses to give enough addresses to you can't (sub)network your home?
The wording is perfectly clear: if you have more than one device, the ISP MUST give you a /64 (under that policy). If you have more than one subnet, the ISP MUST give you a /48.
Yes. I can already see the pricing:
One device: 20$ 2+ devices: 40$ Network of devices: 60$
Would this kind of "voluntary" assignment work in practise? I wouldn't bet a penny on it; ISP's would just do /128 and write the customer contracts so that the other methods would not be possible in practise.
Also, I do not see how a wireless device needs /64 when dial-up wouldn't.
I think it's common today that e.g. 2 home computers share an internet connection. With mobiles and other gadgets going IP this might increase. If /128 assignment is the default, this by default would limit the options people are given. Someone frustrated with different pricing would write a hack to do NATv6. I don't think this should be encouraged. The rules should be set so that the default assignment is at least /64.
-- Pekka Savola "Tell me of difficulties surmounted, Netcore Oy not those you stumble over and fall" Systems. Networks. Security. -- Robert Jordan: A Crown of Swords
-- John WELLS INRIA Rhône-Alpes équipe Planète - ENSIMAG 3A/Télécomm et Réseaux Virginia Tech Networking and Visualization Lab Clé public: finger wells@mansoun.inrialpes.fr <interne>
I could see an ISP giving everyone a /64, but only delegating DNS authority to those /64s who register themselves as networks.
I don't understand why an ISP would be interested in providing a broken service offering to its customers.
I guess it depends on how DNS will pan out in IPv6.
Unless it is administratively easy to provide customer configurable DNS by some automatism, such services are likely not to be offered to mass markets. Personally I would look at this as an engineering challenge: make something that works with ease, rather than limiting the service offering. -hph
Hi, On Wed, May 16, 2001 at 04:11:01PM +0300, Pekka Savola wrote:
Would this kind of "voluntary" assignment work in practise? I wouldn't bet a penny on it; ISP's would just do /128 and write the customer contracts so that the other methods would not be possible in practise.
There is nothing voluntary about this.,
Also, I do not see how a wireless device needs /64 when dial-up wouldn't.
It would, if there's a subnet behind it (bluetooth, whatnot). It wouldn't, otherwise.
I think it's common today that e.g. 2 home computers share an internet connection. With mobiles and other gadgets going IP this might increase.
So they get a /64. No discussion necessary. But for a single-user dynamic IP dialup account ("notebook on the road"), this might just be too much overhead. Gert Doering -- NetMaster -- SpaceNet AG Mail: netmaster@Space.Net Joseph-Dollinger-Bogen 14 Tel : +49-89-32356-0 80807 Muenchen Fax : +49-89-32356-299
Not sure about the large number of lists here... anyway... On Wed, 16 May 2001, Gert Doering wrote:
* IAB/IESG recommended /48. * Use a /128 where it is absolutely known that one and only one device is required, e.g. dialup [<--!!!!!!!] * Use a /64 when sure net will not be subnetted, e.g. a mobile phone given 802.11, bluetooth, etc. ---
I find this thinking, or at least the examples very flawed.
Anyone want to start implementing NATv6 for people whose ISP refuses to give enough addresses to you can't (sub)network your home?
The wording is perfectly clear: if you have more than one device, the ISP MUST give you a /64 (under that policy). If you have more than one subnet, the ISP MUST give you a /48.
I suggest that in this world many ISP's will charge more for the /64 than the /128, and yet more for the /48 than the /64. This means those looking to save money (e.g. students, non-profit organizations, or just cheap people like myself) will probably get the /128, and the next thing you know there's NATng. UltraCheapFastISP User Agreement: "The USER agrees that only a single device will be connected and send data via the PROVIDER's facilities." Personally I think this is fine, but then again I love NAT. :) Is it just me, or does the whole discussion seem academic? Because when many ISP's see the gazillions of addresses they get they're going to allocate it in whatever way they want, and won't bother the RIR's for more space for decades. The only other alternative I see is proactive policing by the RIR's, and I think very few people want that. (I surely don't.) -- Shane "speaking only for myself" Kerr
On Wed, 16 May 2001, Shane Kerr wrote:
I suggest that in this world many ISP's will charge more for the /64 than the /128, and yet more for the /48 than the /64. This means those looking to save money (e.g. students, non-profit organizations, or just cheap people like myself) will probably get the /128, and the next thing you know there's NATng.
To do the above would be plain stupidity, and short sightedness on a grand scale (so it will probably happen!). Here's why: With IPv6 the idea of 1 connected device at home is dead, and so it should be. NAT breaks the internet and it is only fancy programing that fixes it again. This is why IPv6 was designed with e2e in mind. IPv6 isn't just about the internet as we know it today. A lot of companies are looking to use IP networks to provide voice, video, and many other servcies that traditionally are serviced on their own network. This means the home will have just one connection point for all its services. Which in turn means that the home user needs atleast a /64, some may even need a /56. The ISP shouldn't need to charge the home user for the address space for two reasons: 1) There's just so much of it that it is practically worthless. 2) The ISP should be making enough money off the content providers I accept that both these reason won't stop some ISPs from charging, but hopefully it should stop the majority. Either way *all* networks need atleast a /64. Dave "Hoping for a brighter future" Gethings
----- Original Message ----- From: "Shane Kerr" <shane@ripe.net> | UltraCheapFastISP User Agreement: | | "The USER agrees that only a single device will be connected and send | data via the PROVIDER's facilities." | | Personally I think this is fine, but then again I love NAT. :) | | Is it just me, or does the whole discussion seem academic? Because when | many ISP's see the gazillions of addresses they get they're going to | allocate it in whatever way they want, and won't bother the RIR's for | more space for decades. The only other alternative I see is proactive | policing by the RIR's, and I think very few people want that. (I surely | don't.) | | -- | Shane "speaking only for myself" Kerr I sincerely hope so, as this seems to be far from what I understand the current RIPE policies to be. Hans Petter Holen speaking as chair of lir-wg.
participants (5)
-
David Gethings -
Gert Doering -
Hans Petter Holen -
John Wells -
Pekka Savola -
Shane Kerr