IPv6 Pain Experiment

Fri Oct 4 00:34:02 UTC 2019

In article <CAOVYXj-DWX0FhcEJdPMBfec6HcykWtRrErC8wDuBDYYFmE_cWQ at mail.gmail.com> you write:
>that gets me on to my small annoyance... /64 bit subnet masks for
>local networks. really?

Yup.

> ALL of that address space and then throw such
>a large range away on subnets commonly populated
>with no more than a couple of hundred clients...maybe a few thousand
>at worst. what a mistake.

Nope.  The whole point of 128 bit addresses is that you can waste bits
with wild abandon.  My upstream originally assigned me a /64 but since
I have two network segments, they gave me a /48, of which I am using
two /64s.  Since they have a /32, they won't run short of /48's until
they have 65,000 clients with multi segment networks, which will take
a very long time.  In the unlikely event that happens, they can
upgrade their /32 to a /31, since ARIN allocates the /32's with slop
between them.

The programming and configuration is much easier since we can always
assume that every network will have a /64 and no more and no less.

>I come from a background where we had IPv4/DECNET/AppleTalk/IPX all
>around the place - 

Unlike all of them, one mistake that IPv6 did *not* make was to make
addresses too short.

In the same way, IPv6 ULAs are a lot better than IPv4 RFC1918 space.
So long as you follow the spec and pick a truly random ULA prefix,
even if your networks later merge with others the chances of ULAs
colliding rounds to zero.