...another flash-back to the previous millennium below. On 2016-05-19 10:18, Shane Kerr wrote:
Jen and all,
Gather 'round kids, Grandpa has a story!
So this from a Grand-Grand-Pa :-)
At 2016-05-18 20:56:13 +0200 Jen Linkova <furry13@gmail.com> wrote:
On Wed, May 18, 2016 at 8:45 PM, JORDI PALET MARTINEZ [...] I was at a meeting way back in the 20th century - either my first or second time at an IETF, IIRC.
And I remember a late-night meeting amongst the RIR Folks (in Bangkok, iirc), when we tried to come up with a distribution policy which would be a compromise between the different points of view in the 5 regions. It turned out that we were successful in the end. One of the arguments in favour of a uniform prefix length (/48 per 'site') was the support for easy renumbering support in DNS. Anyone remembers the A6 RR type and the level of abstraction/indirection it would have offered? Alas, this beast is extinct by now. RIP...
There was a discussion amongst RIR-types about the default prefix size. Someone proposed the /48 and said that we need to have the same size for every assignment, so that it is easy for customers to move between providers, and to avoid establishing the difference between 'residential customer' and other customers that you are describing. Even at the time I thought it was a bit cheeky to be establishing policy like this, but the idea was not criticized at the meeting. (I also suggested that we don't really need to use 8-bit boundaries even, but was told via some hand-waving arguments about ASICs that this was absolutely necessary.)
I, too, think that this was mostly hand-waving. The only boundary I do consider essential is alignment with the 4bit hex digits of the external representation, in order to have aligned reverse DNS delegation points. FWIW, Wilfried
¯\_(ツ)_/¯
Cheers,
-- Shane