[Lista ArNOG] IPv6 en ISP

[DeLong, Owen]

Forgive my limited Spanish, so I will post in English.

There are many good reasons to stick to /48 instead of /56.

+ Simplified automatic prefix distribution within growing home network
complexity and automation.
+ Improved ability to provide network segmentation in automated topologies
+ Better mechanisms for numbering future home automation and IoT
+ Giving all subscribers the same prefix size simplifies administration and
reduces the potential for human error.

The only possible benefit to selecting /56 is for an ISP to reduce it’s RIR fees.

This is such a minor fraction of costs at any scale where it could be relevant that I think it is both
short-sighted and ill-advised.

Many people are so entrenched in IPv4-think from decades of living in a scarcity mentality that they have
a hard time understanding that so-called conservation in IPv6 is actually wasteful.

Think of it this way… There are two ways to waste addresses:

1. Deploy them in networks where they provide no benefit.
2. Avoid deploying them to such an extent that they are never deployed even up to the
point of protocol extinction.

Obviously in the case of IPv6, we are hoping that confirmation of the second case is many years down the
road and we are hoping that we arrive at the point of protocol extinction for IPv6 with some addresses
still in inventory. However, whether that happens with a single /32 or a dozen /12s or even more really
makes no difference. There’s absolutely no advantage to keeping addresses on the shelf if they are useful.

Consider that IANA has issued roughly 6 /12s from the current /3 designated for IPv6 UNICAST (1 full /12
to each RIR and some various other fragments from earlier experiments and early allocations).

Each /12 contains 2^36 = ~64 Billion /48s. There are 7 Billion people world wide. That means that each
RIR already holds enough address space to grant 9 ISPs enough /48s to allocate one to every single man,
woman, and child on earth and still have lots left over. (There are 5 RIRs, so in fact, worldwide, that’s
45 ISPs).

I don’t know very many people who subscribe to more than 3 or 4 ISPs in total. Even if we assume that
half of that address space is needed for servers, network infrastructure, businesses, etc., we’re still
OK and that’s just in the first 5 /12s. There are 506 times that much space remaining in the IANA free
pool of designated unicast addresses just waiting to go to RIRs if the need arises. That’s
enough addresses for 23,000 ISPs to give EVERY man, woman and child their own /48. (11,500 if we
assume half of all address space goes to infrastructure, servers, businesses, etc.)

So long as we rationally allocate /48s, there will not be an address shortage in IPv6 as a result, so there
is simply no justification and no benefit whatsoever from micro-allocating down to the /56.

Another consideration I will put here:

We don’t know what will happen in the future. We do know what has happened in the past. Today, we have
very screwy and difficult limitations in doing many innovative things on the internet that are a result
of address scarcity and the unfortunate pitfalls of address-sharing technologies like NAT.

Tons of services that would be quick, simple, straightforward, and reliable to implement via a
direct peer-to-peer connection involve convoluted solutions with rendezvous hosts and other hacks
and workarounds just to cope with this address scarcity.

Let’s not encumber the future with the same mistakes of the past.

Some will see that statement as an argument for more aggressive conservation in IPv6, but only if they
have not reviewed the math and the historical details.

When 32 bits was chosen for IPv4, it was not expected to be a direct-to-consumer network. There was
no concept of a “world wide web”. The “browser” did not exist yet. There were no “URLs”. The internet
was something used by researchers and educational institutions and primarily for sending email
and transferring files between mini computers using ASCII terminals and text interfaces.

4 billion seemed like a lot of addresses because we didn’t expect to need to number very much of the
world.

We learned from that mistake and 64 bits was chosen because it would provide enough addresses to
number virtually everything on earth down to the molecule.

(Yes, I said 64 bits… That _WAS_ the initial plan for IPv6, 64 bits with /48 subnets allowing
for 65,536 hosts per subnet).

Later someone decided that stateless address configuration would be useful for lightweight implementations
and wanted to use EUI-64 addresses as an address suffix. This idea appealed to people at the time,
so the address space was extended to 64-bit subnets and to make things fit well in memory chunks
the prefix portion was rounded up to 64 bits as well. While CIDR has been preserved in IPv6 and
all software should fully support any valid prefix length down to /127, conceptually, IPv6 is designed
to be thought of as 64+64 where you have a 64 bit network number and 64 bit host. Further, the
design took into account the growing amount of network equipment in the average home and suggested
the use of /48s as a standard end-site prefix.

They made sure that there would be enough addresses to number EVERYTHING even with such large allocations.

Oh, and there’s also a built-in escape hatch.

See in all of that math I quoted above… That’s just the first 1/8th of the address space. There’s some
special purpose utilization in the last 1/8th as well, so let’s discount that and say that even if we use
up all of those two /3s, we’ve still got 3/4 of the total address space available.

So, I encourage everyone to use /48s and try it this way. If we run out of the first /3 within my lifetime,
I will happily help you work towards more restrictive practices on the remaining 6 /3s.

Owen

On Sep 19, 2016, at 13:39 , Fernando R. Soto <fsoto en fi.uba.ar<mailto:fsoto en fi.uba.ar>> wrote:

Gracias por la respuesta Arturo

Por favor, me explicas pq no alcanzaría un /56 para q el cliente arme sus propias subredes? (256 subredes)
Es un tema de automatización de la creación de esas subredes?


Slds.


De: lista-bounces en arnog.com.ar<mailto:lista-bounces en arnog.com.ar> [mailto:lista-bounces en arnog.com.ar] En nombre de Arturo Servin
Enviado el: lunes, 19 de septiembre de 2016 5:02 p. m.
Para: lista en arnog.com.ar<mailto:lista en arnog.com.ar>
Asunto: Re: [Lista ArNOG] IPv6 en ISP

Fernando

Te vas a quedar corto sobre la hipótesis que los servicios futuros van a requerir mas de una subred. Por ejemplo al dia de hoy podria normal que se requiriera una subred para tus invitados, otra para tu home-networking y otra para tus dispositivos. Porque IPv4 es tan escaso que solo tenemos una, pero si no fuera por eso podriamos tener mas.

Por ello, un /64 o un /56 (256 /64s) sera insuficientes y por ello requieres planear a dar un /48 (aunque ahora des un /56 o /64), de otra forma vas a tener que renumerar y eso te va a costar plata.

Si la hipotesis es valida o no, eso es otra cosa. Lo que si es que te recomendaria no dar /64 y al menos un /56, eso creo que es mas o menos aceptado. Un /48, es debatible aun pero para algunos es lo que recomiendan.

/as

On Mon, 19 Sep 2016 at 19:56 Fernando R. Soto <fsoto en fi.uba.ar<mailto:fsoto en fi.uba.ar>> wrote:
Hola Ariel

No entendí pq me voy a quedar corto. Siendo que entrego un bloque menor a los clientes. /56 en vez de /48

No se podría entregar /64 a clientes hogareños y /56 o /48 a servicios corporativos?

Slds.


De: lista-bounces en arnog.com.ar<mailto:lista-bounces en arnog.com.ar> [mailto:lista-bounces en arnog.com.ar<mailto:lista-bounces en arnog.com.ar>] En nombre de Ariel Weher
Enviado el: lunes, 19 de septiembre de 2016 2:02 p. m.

Para: lista en arnog.com.ar<mailto:lista en arnog.com.ar>
Asunto: Re: [Lista ArNOG] IPv6 en ISP

Con el tiempo te vas a quedar corto. Mi consejo es mantener el /48 y extender el /32 asignado a /29.

No debería ser un problema, porque en v6 abundan las direcciones y no LACNIC va a ser  más permisivo que con v4.

Incluso en RIPE hay políticas aprobadas para entregar /29 en vez de /32, a fin de favorecer algunas técnicas de transición.

S2

2016-09-19 14:00 GMT-03:00 Fernando R. Soto <fsoto en fi.uba.ar<mailto:fsoto en fi.uba.ar>>:
/48 es la escuela de Jordi. Je

A mi q con un solo AS manejo varios Pops se me complicar entregar un /48 por cliente

Q pasa si entrego /56?

Slds.

De: lista-bounces en arnog.com.ar<mailto:lista-bounces en arnog.com.ar> [mailto:lista-bounces en arnog.com.ar<mailto:lista-bounces en arnog.com.ar>] En nombre de Ariel Weher
Enviado el: lunes, 19 de septiembre de 2016 1:39 p. m.
Para: lista en arnog.com.ar<mailto:lista en arnog.com.ar>
Asunto: Re: [Lista ArNOG] IPv6 en ISP

/48 a cada cliente, sin importar que clase de servicio sea.

RFC6177

S2.

2016-09-19 13:10 GMT-03:00 Mariano Carea <mcarea en grupojunin.com.ar<mailto:mcarea en grupojunin.com.ar>>:

Buen día, algún ISP ya implementó IPv6 en sus clientes ? como subdividió la redes ?

gracias

--

Ing. Mariano Carea



Dpto. de Comunicaciones

Grupo Servicios Junín

_______________________________________________
Lista mailing list
Lista en arnog.com.ar<mailto:Lista en arnog.com.ar>
http://mailmancabase.interdotnet.com.ar/mailman/listinfo/lista<https://urldefense.proofpoint.com/v2/url?u=http-3A__mailmancabase.interdotnet.com.ar_mailman_listinfo_lista&d=DQMFaQ&c=96ZbZZcaMF4w0F4jpN6LZg&r=Q_d8tNiSzoBecM8os8iGQA&m=CzEjTBajXfTP9RRXS6wBa3iLtjqvm25nRzQ6XwtBPxE&s=Mf-ttdf2pQ_cJWbLkqxc-ES8syrAXxAuvdyVa7Gi-j0&e=>


_______________________________________________
Lista mailing list
Lista en arnog.com.ar<mailto:Lista en arnog.com.ar>
http://mailmancabase.interdotnet.com.ar/mailman/listinfo/lista<https://urldefense.proofpoint.com/v2/url?u=http-3A__mailmancabase.interdotnet.com.ar_mailman_listinfo_lista&d=DQMFaQ&c=96ZbZZcaMF4w0F4jpN6LZg&r=Q_d8tNiSzoBecM8os8iGQA&m=CzEjTBajXfTP9RRXS6wBa3iLtjqvm25nRzQ6XwtBPxE&s=Mf-ttdf2pQ_cJWbLkqxc-ES8syrAXxAuvdyVa7Gi-j0&e=>

_______________________________________________
Lista mailing list
Lista en arnog.com.ar<mailto:Lista en arnog.com.ar>
http://mailmancabase.interdotnet.com.ar/mailman/listinfo/lista<https://urldefense.proofpoint.com/v2/url?u=http-3A__mailmancabase.interdotnet.com.ar_mailman_listinfo_lista&d=DQMFaQ&c=96ZbZZcaMF4w0F4jpN6LZg&r=Q_d8tNiSzoBecM8os8iGQA&m=CzEjTBajXfTP9RRXS6wBa3iLtjqvm25nRzQ6XwtBPxE&s=Mf-ttdf2pQ_cJWbLkqxc-ES8syrAXxAuvdyVa7Gi-j0&e=>
_______________________________________________
Lista mailing list
Lista en arnog.com.ar<mailto:Lista en arnog.com.ar>
http://mailmancabase.interdotnet.com.ar/mailman/listinfo/lista

------------ próxima parte ------------
Se ha borrado un adjunto en formato HTML...
URL: <http://mailmancabase.interdotnet.com.ar/pipermail/lista/attachments/20160921/05f471b8/attachment-0001.html>