Re: Dual stack IPv6 for IPv4 depletion

[Mark Andrews <marka () isc org>]

In message <559DC43E.5020005 () lugosys com>, "Israel G. Lugo" writes:
> On 07/09/2015 12:59 AM, Mark Andrews wrote:
> > In message <559DB604.8060901 () lugosys com>, "Israel G. Lugo" writes:
> > > Doesn't seem to make sense at all for the ISP side, though. Standard
> > > allocation /32. Giving out /48s. Even if we leave out proper subnet
> > > organization and allocate fully densely, that's at most 65,536 subnets.
> > > Not a very large ISP.
> > /32 is not the standard allocation.  It is the *minimum* allocation
> > for a ISP.  ISPs are expected to ask for *more* addresses to meet their
> > actual requirements.
>
> Thank you for pointing that out. When speaking of /32 I was referring
> specifically to RIPE policy, with which I am more familiar: "Initial
> allocation size" for a LIR is /32, extensive to a /29 with minimal
> bureaucracy. Perhaps I should have said "default allocation".
>
> I understand ISPs should ask for more addresses; however, even e.g. a
> /24 (8x /32) seems to me like it could be "roomier".
>
>
> > > People usually look at IPv6 and focus on the vast numbers of individual
> > > addresses. Naysayers usually get shot down with some quote mentioning
> > > the number of atoms in the universe or some such. Personally, I think
> > > that's a red herring; the real problem is subnets. At this rate I
> > > believe subnets will become the scarce resource sooner or later.
> > No.  People look at /48's for sites.  35,184,372,088,832 /48 sites out of t
> he
> > 1/8th of the total IPv6 space currently in use.  That is 35 trillion sites
> > and if we use that up we can look at using a different default size in the
> > next 1/8th.
> Yes, if we look at end sites individually. My hypothesis is that these
> astronomic numbers are in fact misleading. There isn't, after all, one
> single ISP-Of-The-World, with The-One-Big-Router.
>
> We must divide the addresses by ISPs/LIRs, and so on. Several bits in
> the prefix must be used for subaddressing. A larger ISP will probably
> want to further subdivide its addressing by region, and so on. With
> subdivisions comes "waste". Which is something we don't need to worry
> about at the LAN level, but it would be nice to have that level of
> comfort at the subaddressing level as well.
>
> Let's say I'm a national ISP, using 2001:db8::/32. I divide it like so:
>
> - I reserve 1 bit for future allocation schemes, leaving me a /33;
> - 2 bits for network type (infrastructure, residential, business, LTE): /35
> - 3 bits for geographic region, state, whatever: /38
> - 5 bits for PoP, or city: /43
>
> This leaves me 5 bits for end sites: no joy.
>
> Granted, this is just a silly example, and I don't have to divide my
> address space like this. In fact, I really can't, unless I want to have
> more than 32 customers per city. But I don't think it's a very
> far-fetched example.

A single /48 has enough space/subnets cover the entire infrastructure
of 99.9999% of ISPs even using /64's for p2p links rather than
taking one /64 and subdividing that for all of the p2p links.  Treat
the ISP as a business customer of itself when allocating address
space for itself.  A single /48 or one /48 per site.

> Perhaps I'm missing something obvious here, but it seems to me that it
> would've been nice to have these kinds of possibilities, and more. It
> seems counterintuitive, especially given the "IPv6 way of thinking"
> which is normally encouraged: "stop counting beans, this isn't IPv4".

The thinking is that ISP's are experts and can deal with managing
complex allocation policy.  They can also deal with multiple more
specific routes etc.  They already cope with this in IPv4.

> Regards,
> Israel G. Lugo
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka () isc org