Orbit

Hi Skeeve,

Could you please relay my query to the APNIC-TALK list ? I dont have the email address for them.

I am willing to participate on any current forum which discusses the address assignment recommendations.

In my opinion, stateless autoconfiguration is little justification (to start assigning /64s for end-customers) when it comes to the issue of address exhaustion which was the main driver to come up with a 128 bit address space. If we start assigning /64s to end-customers right from day-zero, we are effectively halving the whole 128 bit address space and making all the vital /64s unavailable for use in future.

We can always ask vendors to modify the stateless autoconfiguration algorithm and look into slightly more conservative addressing scheme.

And we all know how difficult it is to reclaim address spaces from customers once they have deployed them.

We are talking about assigning 2^64 addresses to potentially small-scale customers ?? I don't understand the justification there.

Randy Bush: Yes I have read RFC 6177 and I am mainly concerned about its recommendations of assigning /64s - some ISPs in Australia are now taking these recommendations and assigning even residential edge customers with a /64 IPv6 space (I found this out after participating in AUS-NOG conference and was alarmed by this).

I can be reached on the following:

Usman Latif Senior Network Engineer Uecomm (Singtel-Optus Limited) Phone: +61 2 8085 3212 Email: ulatif@uecomm.com.au Sydney, Australia

regards Usman Latif

________________________________ From: Skeeve Stevens Skeeve@eintellego.net To: Usman Latif osmankh@yahoo.com; "sig-policy@lists.apnic.net" sig-policy@lists.apnic.net Sent: Friday, 16 September 2011 6:49 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

Hi Usman,

This is a good question and worth discussing.  But, it should be discussed in perhaps the apnic-talk list, not the sig-policy list, which is for policy related discussion.

Let's take it over there, and let the discussion begin!

…Skeeve

APNIC Sig-Policy Co-Chair   -- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call- Juniper - HP Networking - Cisco - Brocade - Huawei On 16/09/11 6:09 PM, "Usman Latif" osmankh@yahoo.com wrote:

Hi,

I am trying to understand the reasoning and logic behind IETF/IANA's decision to recommend assignments of /64 addresses to residential CPEs ?? In my opinion, this would result in a lot of unnecessary wastage of IPv6 address space.

Can someone help me to point towards the drivers behind this thinking?

IMO a /96 IPv6 assignment to residential customers is more than enough.

regards Usman

* sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

Hi Randy,

I have done binary math for long enough to know what I am saying :) but I acknowledge that may be I worded it incorrectly in my last email.

The point I am trying to make here is that a /64 seems just too big of an address space to assign to end-customers (especially when these end-customers donot happen to be enterprises or corporate customers but SOHOs or residential edge customers).

regards Usman

________________________________ From: Randy Bush randy@psg.com To: Usman Latif osmankh@yahoo.com Cc: Skeeve Stevens Skeeve@eintellego.net; "sig-policy@lists.apnic.net" sig-policy@lists.apnic.net Sent: Friday, 16 September 2011 8:07 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

I am willing to participate on any current forum which discusses the address assignment recommendations.

you're there

If we start assigning /64s to end-customers right from day-zero, we are effectively halving the whole 128 bit address space

your math is bad

randy

Hey Usman,

As you were just at the AUSNOG conference, if you posted on the AUSNOG list you may get the opinions from those at the conference you attended – and why they are choosing to do it that way.

I think the primary driver of /64 is to support EUI64, and as such most are going longer – to a 56, or perhaps a 48 to allow for multiple networks at the same site… although it does feel like an awful lot of space, it seems reasonable given the availability of the space and the preference to allow SLAAC to function.

...Skeeve

--

Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists

skeeve@eintellego.net ; www.eintellego.net

Phone: 1300 753 383 ; Fax: (+612) 8572 9954

Cell +61 (0)414 753 383 ; skype://skeeve

facebook.com/eintellego or eintellego@facebook.com

twitter.com/networkceoau ; www.linkedin.com/in/skeeve

PO Box 7726, Baulkham Hills, NSW 1755 Australia

--

eintellego - The Experts that the Experts call

- Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 7:56 PM, "Usman Latif" <osmankh@yahoo.commailto:osmankh@yahoo.com> wrote:

Hi Skeeve,

Could you please relay my query to the APNIC-TALK list ? I dont have the email address for them. I am willing to participate on any current forum which discusses the address assignment recommendations.

In my opinion, stateless autoconfiguration is little justification (to start assigning /64s for end-customers) when it comes to the issue of address exhaustion which was the main driver to come up with a 128 bit address space. If we start assigning /64s to end-customers right from day-zero, we are effectively halving the whole 128 bit address space and making all the vital /64s unavailable for use in future.

We can always ask vendors to modify the stateless autoconfiguration algorithm and look into slightly more conservative addressing scheme.

And we all know how difficult it is to reclaim address spaces from customers once they have deployed them.

We are talking about assigning 2^64 addresses to potentially small-scale customers ?? I don't understand the justification there.

Randy Bush: Yes I have read RFC 6177 and I am mainly concerned about its recommendations of assigning /64s - some ISPs in Australia are now taking these recommendations and assigning even residential edge customers with a /64 IPv6 space (I found this out after participating in AUS-NOG conference and was alarmed by this).

I can be reached on the following:

Usman Latif Senior Network Engineer Uecomm (Singtel-Optus Limited) Phone: +61 2 8085 3212 Email: ulatif@uecomm.com.aumailto:ulatif@uecomm.com.au Sydney, Australia

regards Usman Latif

________________________________ From: Skeeve Stevens <Skeeve@eintellego.netmailto:Skeeve@eintellego.net> To: Usman Latif <osmankh@yahoo.commailto:osmankh@yahoo.com>; "sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net" <sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net> Sent: Friday, 16 September 2011 6:49 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

Hi Usman,

This is a good question and worth discussing. But, it should be discussed in perhaps the apnic-talk list, not the sig-policy list, which is for policy related discussion.

Let's take it over there, and let the discussion begin!

…Skeeve

APNIC Sig-Policy Co-Chair

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.netmailto:skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.commailto:eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call - Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 6:09 PM, "Usman Latif" <osmankh@yahoo.commailto:osmankh@yahoo.com> wrote:

Hi,

I am trying to understand the reasoning and logic behind IETF/IANA's decision to recommend assignments of /64 addresses to residential CPEs ?? In my opinion, this would result in a lot of unnecessary wastage of IPv6 address space.

Can someone help me to point towards the drivers behind this thinking?

IMO a /96 IPv6 assignment to residential customers is more than enough.

regards Usman * sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

So it seems that a previous RFC (regarding stateless autoconfig) is driving us to use /64s for edge-assignments? - this to me seems more like a constrain than a reason.

I just hope that we don't look back at this time in future and regret taking this decision because it seems that we are starting out very very liberally in our address assignment approach (potentially wasting a lot of space) and could potentially come to a similar exhaustion problem far earlier than if we had started out more conservatively with /96s or something similar.

In all honesty, I am not convinced that we have the proper justification or a reason to suggest using /64s for edge-assignments on day-zero of IPv6 rollout in the world.

regards Usman

________________________________ From: Skeeve Stevens Skeeve@eintellego.net To: Skeeve Stevens Skeeve@eintellego.net; Usman Latif osmankh@yahoo.com; "sig-policy@lists.apnic.net" sig-policy@lists.apnic.net Sent: Friday, 16 September 2011 8:47 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

I seem to have just repeated most of what AJ just said.

...Skeeve   -- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call- Juniper - HP Networking - Cisco - Brocade - Huawei On 16/09/11 8:46 PM, "Skeeve Stevens" Skeeve@eintellego.net wrote:

Hey Usman,

As you were just at the AUSNOG conference, if you posted on the AUSNOG list you may get the opinions from those at the conference you attended – and why they are choosing to do it that way.

I think the primary driver of /64 is to support EUI64, and as such most are going longer – to a 56, or perhaps a 48 to allow for multiple networks at the same site… although it does feel like an awful lot of space, it seems reasonable given the availability of the space and the preference to allow SLAAC to function.

...Skeeve   -- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call- Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 7:56 PM, "Usman Latif" osmankh@yahoo.com wrote:

Hi Skeeve,

...

Could you please relay my query to the APNIC-TALK list ? I dont have the email address for them.

I am willing to participate on any current forum which discusses the address assignment recommendations.

In my opinion, stateless autoconfiguration is little justification (to start assigning /64s for end-customers) when it comes to the issue of address exhaustion which was the main driver to come up with a 128 bit address space. If we start assigning /64s to

end-customers right from day-zero, we are effectively halving the whole 128 bit address space and making all the vital /64s unavailable for use in future.

...

We can always ask vendors to modify the stateless autoconfiguration algorithm and look into slightly more conservative addressing scheme.

And we all know how difficult it is to reclaim address spaces from customers once they have deployed them.

We are talking about assigning 2^64 addresses to potentially small-scale customers ?? I don't understand the justification there.

Randy Bush: Yes I have read RFC 6177 and I am mainly concerned about its recommendations of assigning /64s - some ISPs in Australia are now taking these recommendations and assigning even residential edge customers with a /64 IPv6 space (I found this out after participating in AUS-NOG conference and was alarmed by this).

I can be reached on the following:

Usman Latif Senior Network Engineer Uecomm (Singtel-Optus Limited) Phone: +61 2 8085 3212 Email: ulatif@uecomm.com.au Sydney, Australia

regards Usman Latif

---

From: Skeeve Stevens Skeeve@eintellego.net To: Usman Latif osmankh@yahoo.com; "sig-policy@lists.apnic.net" sig-policy@lists.apnic.net Sent: Friday, 16 September 2011 6:49 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

Hi Usman,

This is a good question and worth discussing.  But, it should be discussed in perhaps the apnic-talk list, not the sig-policy list, which is for policy related discussion.

Let's take it over there, and let the discussion begin!

…Skeeve

APNIC Sig-Policy Co-Chair   -- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call- Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 6:09 PM, "Usman Latif" osmankh@yahoo.com wrote:

Hi,

...

I am trying to understand the reasoning and logic behind IETF/IANA's decision to recommend assignments of /64 addresses to residential CPEs ?? In my opinion, this would result in a lot of unnecessary wastage of IPv6 address space.

Can someone help me to point towards the drivers behind this thinking?

IMO a /96 IPv6 assignment to residential customers is more than enough.

regards Usman

* sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

...

* sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

i am fairly sure all your competitors would admire your parsimony and cheer your issuing your customers only a /96.

randy

Gents,

While I agree that autoconf is a major constraint and if I think selfishly about being liberal and competitive in the market, I shouldn't even care about assigning a larger subnet to residential CPEs.

But my perspective on this issue is without regards to businesses or ISPs or profits - its simply about whether we are wasting too many addresses for scenarios where, at the max, customers would probably never go beyond few thousand IPs ever. And whether we are likely to hit the same address exhaustion (as we face with IPv4 today) due to this liberal assignment.

And I seriously doubt that if we don't take the right decision about assignments to residential CPEs, it would be almost impossible to reclaim address spaces in future should we hit the exhaustion again.

We also know how at the start of IPv4, nobody could possibly imagine chewing up all 4 billion addresses and I fear some people are having the same mindset that having 2^64 amount of /64s out there are more than enough till the end of time :).

While its fine to assign a /64 to enterprises and corporate customers LANs, using the same technique in residential or high-volume single-user customer scenarios is gonna result in simple wastage of almost ~ 2^64 = 18,446,744,073,709,551,616 IPv6 addresses that could otherwise have been used where needed. I see this as a big waste of address space.

Autoconf of end-user addresses should not be considered as a driver for using /64s but a constraint that is making us take the decision of assigning /64 level subnets to end-users.

Anyway, if most of the community does not see from my perspective, I have no choice but to agree with majority's decision on this matter :) but I rest my case with everyone out there and have provided my 2 cents on this matter.

regards Usman

________________________________ From: Alastair Johnson aj@sneep.net To: Randy Bush randy@psg.com; sig-policy-bounces@lists.apnic.net; Usman Latif osmankh@yahoo.com Cc: APNIC Address Policy SIG sig-policy@apnic.net Sent: Saturday, 17 September 2011 2:31 AM Subject: Re: [sig-policy] Need to understand logic behind assigning /64IPv6 addresses

Indeed - I was going to make that point too.

*not* using /64s (or shorter) is going to make your helpdesk and customers unhappy... But as Randy says, your competitors will be happy.

-----Original Message----- From: Randy Bush randy@psg.com Sender: sig-policy-bounces@lists.apnic.net Date: Fri, 16 Sep 2011 17:59:46 To: Usman Latifosmankh@yahoo.com Cc: APNIC Address Policy SIGsig-policy@apnic.net Subject: Re: [sig-policy] Need to understand logic behind assigning /64     IPv6    addresses

i am fairly sure all your competitors would admire your parsimony and cheer your issuing your customers only a /96.

randy *              sig-policy:  APNIC SIG on resource management policy          * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

Thanks for understanding.

To me it just seems like a crazy idea of assigning a /64 subnet (that can otherwise fulfill requirements of 18,446,744,073,709,551,616 hosts) to a residential CPE which today uses at the max 5-10 IPv4 addresses on their LANs.

I think because we have been working with IPv4 and have been really careful about not wasting IPs due to the limited address space, maybe its that same mindset which makes me uncomfortable looking at multiple /64s going to a single DSL end-customer etc. :)

Anyway, next time I am using autoconf and assign a /64 to a residential CPE, I'll just go blind and convince myself not to think about wasting all these addresses....

________________________________ From: Randy Bush randy@psg.com To: Usman Latif osmankh@yahoo.com Cc: APNIC Address Policy SIG sig-policy@apnic.net Sent: Saturday, 17 September 2011 4:42 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64IPv6 addresses

it is not that i disagree with your arguments.  it's just that i lost the selfsame arguments some years ago.  the current generally accepted 'wisdom,' with little supporting measurement, is that iv6 address space is effectively infinite and the ipv6 routing table is a very scarce resource.  that this resembles pigs at a trough constructing a barrier to entry is not a widely held perception.

randy

On Sep 17, 2011, at 3:21 AM, Usman Latif wrote:

Thanks for understanding.

To me it just seems like a crazy idea of assigning a /64 subnet (that can otherwise fulfill requirements of 18,446,744,073,709,551,616 hosts) to a residential CPE which today uses at the max 5-10 IPv4 addresses on their LANs.

Assigning a /64 is crazy... You should not be assigning your residential customers only a single subnet. Ideally, they should receive a /48, but, at an absolute minimum, I would think a /56.

It's not about counting hosts. IPv6 was designed around the idea that counting hosts should be unnecessary and the 64 bits you are now complaining about wasting were added for the purpose of using them in this fashion. If they hadn't been allocated to this purpose, likely IPv6 would have been a 64 bit address rather than a 128 bit address.

I think because we have been working with IPv4 and have been really careful about not wasting IPs due to the limited address space, maybe its that same mindset which makes me uncomfortable looking at multiple /64s going to a single DSL end-customer etc. :)

Yes... You are most definitely suffering from IPv4-think mindset. To deploy IPv6 properly and not in a manner which will detract from the development of better end user technologies, you really need to move beyond the scarcity mindset. As I have said several times. let's try allocating IPv6 as it was intended (/48 per end site regardless of whether it's residential, commercial, etc.) and if we use up even so much as 20 /12s in less than 50 years, I will accept that we need to consider more conservative allocation strategies.

This won't create a need to reclaim. The safety valve I am suggesting (at 20 /12s) leaves us with more than 3,564 /12s still in reserve to use with a more conservative allocation policy. (that without invading c000::/3 which is where multicast, link local, etc. are all reserved).

Anyway, next time I am using autoconf and assign a /64 to a residential CPE, I'll just go blind and convince myself not to think about wasting all these addresses....

You shouldn't have to go blind. You should be able to do it by simply recognizing that the design of the protocol is different in this way and that those 64 bits were added for that purpose.

Owen

From: Randy Bush randy@psg.com To: Usman Latif osmankh@yahoo.com Cc: APNIC Address Policy SIG sig-policy@apnic.net Sent: Saturday, 17 September 2011 4:42 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64IPv6 addresses

it is not that i disagree with your arguments. it's just that i lost the selfsame arguments some years ago. the current generally accepted 'wisdom,' with little supporting measurement, is that iv6 address space is effectively infinite and the ipv6 routing table is a very scarce resource. that this resembles pigs at a trough constructing a barrier to entry is not a widely held perception.

randy

•          sig-policy:  APNIC SIG on resource management policy           *
  

---

sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

Hey Usman,

Thing is… I agree with you when it comes to Interconnects, and so on, where I am much more conservative, using /112's where a lot of people are using much larger (64s, etc).

But when it comes to end user assignment…. We're not really talking about wastage here.

At a minimum, service providers are allocated a /32. A single /32 contains 4.2billion /64's (I think my math is correct).

If that is the case… and Uecomm has a /32, I don't think you are going to have a problem for a long time if you use /64's. This is a key reason that many are handing out /56's…. There is 16.7million /56's in a /32.

So while it might seem like wastage… giving a /64, or a /56 to each end user, really, in the end is meaningless in the scale of things…

The Australian Defence Department has a /20… which is 4096 /32's… that is one organisation, and I know Telstra has at least a /20 plus some others. I don't imagine either of them running out in many years – even if they assigned a /64 or /56 to every… thing they have.

...Skeeve

--

Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists

skeeve@eintellego.net ; www.eintellego.net

Phone: 1300 753 383 ; Fax: (+612) 8572 9954

Cell +61 (0)414 753 383 ; skype://skeeve

facebook.com/eintellego or eintellego@facebook.com

twitter.com/networkceoau ; www.linkedin.com/in/skeeve

PO Box 7726, Baulkham Hills, NSW 1755 Australia

--

eintellego - The Experts that the Experts call

- Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 9:02 PM, "Usman Latif" <osmankh@yahoo.commailto:osmankh@yahoo.com> wrote:

So it seems that a previous RFC (regarding stateless autoconfig) is driving us to use /64s for edge-assignments? - this to me seems more like a constrain than a reason.

I just hope that we don't look back at this time in future and regret taking this decision because it seems that we are starting out very very liberally in our address assignment approach (potentially wasting a lot of space) and could potentially come to a similar exhaustion problem far earlier than if we had started out more conservatively with /96s or something similar.

In all honesty, I am not convinced that we have the proper justification or a reason to suggest using /64s for edge-assignments on day-zero of IPv6 rollout in the world.

regards Usman

________________________________ From: Skeeve Stevens <Skeeve@eintellego.netmailto:Skeeve@eintellego.net> To: Skeeve Stevens <Skeeve@eintellego.netmailto:Skeeve@eintellego.net>; Usman Latif <osmankh@yahoo.commailto:osmankh@yahoo.com>; "sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net" <sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net> Sent: Friday, 16 September 2011 8:47 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

I seem to have just repeated most of what AJ just said.

...Skeeve

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.netmailto:skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.commailto:eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call - Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 8:46 PM, "Skeeve Stevens" <Skeeve@eintellego.netmailto:Skeeve@eintellego.net> wrote:

Hey Usman,

As you were just at the AUSNOG conference, if you posted on the AUSNOG list you may get the opinions from those at the conference you attended – and why they are choosing to do it that way.

I think the primary driver of /64 is to support EUI64, and as such most are going longer – to a 56, or perhaps a 48 to allow for multiple networks at the same site… although it does feel like an awful lot of space, it seems reasonable given the availability of the space and the preference to allow SLAAC to function.

...Skeeve

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.netmailto:skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.commailto:eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call - Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 7:56 PM, "Usman Latif" <osmankh@yahoo.commailto:osmankh@yahoo.com> wrote:

Hi Skeeve,

Could you please relay my query to the APNIC-TALK list ? I dont have the email address for them. I am willing to participate on any current forum which discusses the address assignment recommendations.

In my opinion, stateless autoconfiguration is little justification (to start assigning /64s for end-customers) when it comes to the issue of address exhaustion which was the main driver to come up with a 128 bit address space. If we start assigning /64s to end-customers right from day-zero, we are effectively halving the whole 128 bit address space and making all the vital /64s unavailable for use in future.

We can always ask vendors to modify the stateless autoconfiguration algorithm and look into slightly more conservative addressing scheme.

And we all know how difficult it is to reclaim address spaces from customers once they have deployed them.

We are talking about assigning 2^64 addresses to potentially small-scale customers ?? I don't understand the justification there.

Randy Bush: Yes I have read RFC 6177 and I am mainly concerned about its recommendations of assigning /64s - some ISPs in Australia are now taking these recommendations and assigning even residential edge customers with a /64 IPv6 space (I found this out after participating in AUS-NOG conference and was alarmed by this).

I can be reached on the following:

Usman Latif Senior Network Engineer Uecomm (Singtel-Optus Limited) Phone: +61 2 8085 3212 Email: ulatif@uecomm.com.aumailto:ulatif@uecomm.com.au Sydney, Australia

regards Usman Latif

________________________________ From: Skeeve Stevens <Skeeve@eintellego.netmailto:Skeeve@eintellego.net> To: Usman Latif <osmankh@yahoo.commailto:osmankh@yahoo.com>; "sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net" <sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net> Sent: Friday, 16 September 2011 6:49 PM Subject: Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

Hi Usman,

This is a good question and worth discussing. But, it should be discussed in perhaps the apnic-talk list, not the sig-policy list, which is for policy related discussion.

Let's take it over there, and let the discussion begin!

…Skeeve

APNIC Sig-Policy Co-Chair

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.netmailto:skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.commailto:eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call - Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 6:09 PM, "Usman Latif" <osmankh@yahoo.commailto:osmankh@yahoo.com> wrote:

Hi,

I am trying to understand the reasoning and logic behind IETF/IANA's decision to recommend assignments of /64 addresses to residential CPEs ?? In my opinion, this would result in a lot of unnecessary wastage of IPv6 address space.

Can someone help me to point towards the drivers behind this thinking?

IMO a /96 IPv6 assignment to residential customers is more than enough.

regards Usman * sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

* sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

* sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

I agree with you when it comes to Interconnects, and so on, where I am much more conservative, using /112's where a lot of people are using much larger (64s, etc).

you may want to look at rfc 5952

randy

Randy,

While you bring up RFC6164, a quick question for you:

Why in the end was it decided that RFC6164 didn't update RFC4291?

At the moment, if I were to be very strict in interpretation (which I'm not intending to be) then I think this means that RFC6164 only requires that all routers support /127, but that only /64s are still formally assignable, seeing I don't think RFC6164 actually says "it is acceptable within IPv6 addressing architecture to assign /127 for point-to-point links" and it doesn't fornally update RFC4291.

(As I said, I'm not intending to be that strict, but of course the other reason for not using /127s at the moment is that some equipment hasn't implemented RFC6164 yet, which means there can be difficulties using it right now.)

Regards, David

Sent from my iPhone

On 17/09/2011, at 2:50 AM, Randy Bush randy@psg.com wrote:

...

...

I agree with you when it comes to Interconnects, and so on, where I am much more conservative, using /112's where a lot of people are using much larger (64s, etc).

you may want to look at rfc 5952

aiiii! sorry. rfc 6164

randy

•          sig-policy:  APNIC SIG on resource management  
  

policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

Hi, Thanks for sharing calculations. On Fri, Sep 16, 2011 at 7:37 PM, Skeeve Stevens Skeeve@eintellego.netwrote:

Hey Usman,

Thing is… I agree with you when it comes to Interconnects, and so on, where I am much more conservative, using /112's where a lot of people are using much larger (64s, etc).

But when it comes to end user assignment…. We're not really talking about wastage here.

At a minimum, service providers are allocated a /32. A single /32 contains 4.2billion /64's (I think my math is correct).

If that is the case… and Uecomm has a /32, I don't think you are going to have a problem for a long time if you use /64's. This is a key reason that many are handing out /56's…. There is 16.7million /56's in a /32.

So while it might seem like wastage… giving a /64, or a /56 to each end user, really, in the end is meaningless in the scale of things…

The Australian Defence Department has a /20… which is 4096 /32's… that is one organisation, and I know Telstra has at least a /20 plus some others. I don't imagine either of them running out in many years – even if they assigned a /64 or /56 to every… thing they have.

APNIC may have received justification for allocating these blocks from Australian Defense and Telstra.Can justification given by them be shared on this list. Just want to understand,how can they use some many IP addresses.

...Skeeve

--

Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists

skeeve@eintellego.net ; www.eintellego.net

Phone: 1300 753 383 ; Fax: (+612) 8572 9954

Cell +61 (0)414 753 383 ; skype://skeeve

facebook.com/eintellego or eintellego@facebook.com

twitter.com/networkceoau ; www.linkedin.com/in/skeeve

PO Box 7726, Baulkham Hills, NSW 1755 Australia

--

eintellego - The Experts that the Experts call

• Juniper - HP Networking - Cisco - Brocade - Huawei

  On 16/09/11 9:02 PM, "Usman Latif" osmankh@yahoo.com wrote:

  So it seems that a previous RFC (regarding stateless autoconfig) is

driving us to use /64s for edge-assignments? - this to me seems more like a constrain than a reason.

I just hope that we don't look back at this time in future and regret taking this decision because it seems that we are starting out very very liberally in our address assignment approach (potentially wasting a lot of space) and could potentially come to a similar exhaustion problem far earlier than if we had started out more conservatively with /96s or something similar.

In all honesty, I am not convinced that we have the proper justification or a reason to suggest using /64s for edge-assignments on day-zero of IPv6 rollout in the world.

regards Usman

---

*From:* Skeeve Stevens Skeeve@eintellego.net *To:* Skeeve Stevens Skeeve@eintellego.net; Usman Latif < osmankh@yahoo.com>; "sig-policy@lists.apnic.net" < sig-policy@lists.apnic.net> *Sent:* Friday, 16 September 2011 8:47 PM *Subject:* Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

I seem to have just repeated most of what AJ just said.

...Skeeve

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call

• Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 8:46 PM, "Skeeve Stevens" Skeeve@eintellego.net wrote:

Hey Usman,

As you were just at the AUSNOG conference, if you posted on the AUSNOG list you may get the opinions from those at the conference you attended – and why they are choosing to do it that way.

I think the primary driver of /64 is to support EUI64, and as such most are going longer – to a 56, or perhaps a 48 to allow for multiple networks at the same site… although it does feel like an awful lot of space, it seems reasonable given the availability of the space and the preference to allow SLAAC to function.

...Skeeve

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call

• Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 7:56 PM, "Usman Latif" osmankh@yahoo.com wrote:

Hi Skeeve,

Could you please relay my query to the APNIC-TALK list ? I dont have the email address for them. I am willing to participate on any current forum which discusses the address assignment recommendations.

In my opinion, stateless autoconfiguration is little justification (to start assigning /64s for end-customers) when it comes to the issue of address exhaustion which was the main driver to come up with a 128 bit address space. If we start assigning /64s to end-customers right from day-zero, we are effectively halving the whole 128 bit address space and making all the vital /64s unavailable for use in future.

We can always ask vendors to modify the stateless autoconfiguration algorithm and look into slightly more conservative addressing scheme.

And we all know how difficult it is to reclaim address spaces from customers once they have deployed them.

We are talking about assigning 2^64 addresses to potentially small-scale customers ?? I don't understand the justification there.

Randy Bush: Yes I have read RFC 6177 and I am mainly concerned about its recommendations of assigning /64s - some ISPs in Australia are now taking these recommendations and assigning even residential edge customers with a /64 IPv6 space (I found this out after participating in AUS-NOG conference and was alarmed by this).

I can be reached on the following:

Usman Latif Senior Network Engineer Uecomm (Singtel-Optus Limited) Phone: +61 2 8085 3212 Email: ulatif@uecomm.com.au Sydney, Australia

regards Usman Latif

---

*From:* Skeeve Stevens Skeeve@eintellego.net *To:* Usman Latif osmankh@yahoo.com; "sig-policy@lists.apnic.net" < sig-policy@lists.apnic.net> *Sent:* Friday, 16 September 2011 6:49 PM *Subject:* Re: [sig-policy] Need to understand logic behind assigning /64 IPv6 addresses

Hi Usman,

This is a good question and worth discussing. But, it should be discussed in perhaps the apnic-talk list, not the sig-policy list, which is for policy related discussion.

Let's take it over there, and let the discussion begin!

…Skeeve

APNIC Sig-Policy Co-Chair

-- Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists skeeve@eintellego.net ; www.eintellego.net Phone: 1300 753 383 ; Fax: (+612) 8572 9954 Cell +61 (0)414 753 383 ; skype://skeeve facebook.com/eintellego or eintellego@facebook.com twitter.com/networkceoau ; www.linkedin.com/in/skeeve PO Box 7726, Baulkham Hills, NSW 1755 Australia

-- eintellego - The Experts that the Experts call

• Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 6:09 PM, "Usman Latif" osmankh@yahoo.com wrote:

Hi,

I am trying to understand the reasoning and logic behind IETF/IANA's decision to recommend assignments of /64 addresses to residential CPEs ?? In my opinion, this would result in a lot of unnecessary wastage of IPv6 address space.

Can someone help me to point towards the drivers behind this thinking?

IMO a /96 IPv6 assignment to residential customers is more than enough.

regards Usman

• sig-policy: APNIC SIG on resource management policy *

_______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

* sig-policy: APNIC SIG on resource management policy *

_______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

• sig-policy: APNIC SIG on resource management policy *

_______________________________________________ sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

•          sig-policy:  APNIC SIG on resource management policy
  

---

sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

--

Hi Ajay and all,

Leo is correct. Justification of resource request is classified as confidential material. The APNIC membership agreement (section 3.1.g) prohibits us from disclosing any confidential information that a Member provides to the APNIC Secretariat.

Please also note that justification required and delegation size will strictly follow the applicable policies at the time of the request.

Regards, ________________________________________________________________________ Sanjaya email: sanjaya@apnic.net Services Director, APNIC sip: sanjaya@voip.apnic.net http://www.apnic.net phone: +61 7 3858 3100 ________________________________________________________________________ * Sent by email to save paper. Print only if necessary.

On 17/09/2011 12:52 PM, Leo Vegoda wrote:

Ajay,

On Sep 16, 2011, at 7:28 PM, "Ajay Kumar" <joinajay1@gmail.com mailto:joinajay1@gmail.com> wrote:

[...]

...

APNIC may have received justification for allocating these blocks from Australian Defense and Telstra.Can justification given by them be shared on this list. Just want to understand,how can they use some many IP addresses.

The people who received the allocations might but APNIC won't because its policy doesn't allow it to share justifications:

"6.2. Security and confidentiality

APNIC will maintain systems and practices that protect the confidentiality of all information relating to the commercial and infrastructure operations of all Members and their customers. APNIC will ensure that the employment of all of its staff or agents is based upon an explicit condition of confidentiality regarding such information."

http://www.apnic.net/policy/policy-environment#security

Regards,

Leo

•          sig-policy:  APNIC SIG on resource management policy           *
  

---

sig-policy mailing list sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

Usman,

On Sep 16, 2011, at 4:02 AM, Usman Latif wrote:

I just hope that we don't look back at this time in future and regret taking this decision because it seems that we are starting out very very liberally in our address assignment approach (potentially wasting a lot of space) and could potentially come to a similar exhaustion problem far earlier than if we had started out more conservatively with /96s or something similar.

Do the math.

Last I looked, we were consuming around 300 million IPv4 addresses per year.

Assume we allocate 300 million IPv6 /64s per year. It would take 7,686,143,364 years to consume the 1/8th of the IPv6 address space used for global unicast. If we assume we allocate 300 million IPv6 /48s per year (as recommended by the IETF), we run out of the current global unicast block in 117,281 years.

Let's assume the IPv6 Internet grows 1000 times faster than the IPv4 Internet. If we count /48s as the minimum allocation unit, we'll run out of the current global unicast block in 117 years (7,686,143 years if we count /64s).

And once we consume the current global unicast block, we still have 7/8th of the address space left. If we consume the first 1/8th in my lifetime, I will think it appropriate that everyone involved in devising allocation policy be taken out back and be beaten with a stick.

Under current policies, there are far more useful things to worry about than running out of IPv6 space. The real risk is that policies will change since there is no finite resource that policy wonks can't devise insane policies to consume.

Regards, -drc

Sorry Randy,

My assumption was this list was to discuss actual policy as it pertains to the policy process of APNIC, not theory and general discussion, which would probably get more input on a wider lists like apnic-talk, ausnog, nanog, etc. If the list turns into a generic discussion list with a high level of non-policy related traffic, it would become a challenge for the Chairs to assemble the views of the list as it related to actual policy.

...Skeeve

--

Skeeve Stevens, CEO - eintellego Pty Ltd - The Networking Specialists

skeeve@eintellego.net ; www.eintellego.net

Phone: 1300 753 383 ; Fax: (+612) 8572 9954

Cell +61 (0)414 753 383 ; skype://skeeve

facebook.com/eintellego or eintellego@facebook.com

twitter.com/networkceoau ; www.linkedin.com/in/skeeve

PO Box 7726, Baulkham Hills, NSW 1755 Australia

--

eintellego - The Experts that the Experts call

- Juniper - HP Networking - Cisco - Brocade - Huawei

On 16/09/11 7:28 PM, "Randy Bush" <randy@psg.commailto:randy@psg.com> wrote:

I am trying to understand the reasoning and logic behind IETF/IANA's decision

you have contradictory words in the same sentence, iana and logic.

could you point out ietf _recommending_ a /64 for residential customers? to the best of my knowledge, the current ietf thinking is best codified in RFC 6177.

IMO a /96 IPv6 assignment to residential customers is more than enough.

the downside of this would be that the customer could not use auto-conf on their lan. the general religion on the subject is that ipv6 space is effectively infinite (a tenet to which i do not subscribe), and a very large allocation, e.g. /48 or shorter, should be assigned to each customer.

/56 and /48 are commonly used sizes.

and this is a perfectly good list on which to discuss this aspect of addressing policy.

randy * sig-policy: APNIC SIG on resource management policy * _______________________________________________ sig-policy mailing list sig-policy@lists.apnic.netmailto:sig-policy@lists.apnic.net http://mailman.apnic.net/mailman/listinfo/sig-policy

A personal thought on this topic:

I don't actually mind the /64 basis, for the reasons that have been discussed by others. But I do wonder about the industry trend towards /56 per broadband service.

If I consider /64 as representing a physical or logical network segment (as essentially implied by RFC4291), then that means each broadband service with a /56 will have the capability to support 256 virtual or physical networks in the home. Even allowing in principle for different service sets to each have their own vlan, 256 vlans still seems a little excessive for the average home - most of which will probably only be using one network for a long time to come.

I would have thought that 16 vlans - i.e. a /60 - would be a good standard number to start with, especially if you assume dynamic prefix delegation would mean that you could readily upgrade a customer's address range to a shorter prefix if it really proved necessary.

Also, the implications on the number of customers that can be supported per /32 are significant; with /56 per service, you can only support 17M customers with a /32, but with /60 per service, that becomes 268M.

I agree with all the previous statements that this isn't going to unduly exhaust the IPv6 space any time soon, but like some others I wonder whether we are now tending towards excessive allocation sizes - and for me this is specifically about consumer services. As I said, I don't mind the basic /64 per network, and at the other end I don't mind a /48 per enterprise site, because I suspect the number of such sites is many orders of magnitude smaller than the numbers of consumers or small businesses. But 256 lans per domestic premise does make me wonder...

Having said all this, I likewise feel that the industry has made up its mind and we probably need to be consistent to avoid confusion while we're trying to get IPv6 out there.

But I would be interested in any other thoughts on this - would anyone else care to share their views as to whether /56 per broadband service seems too much (or too little) or just right?

Regards,

David Woodgate