nanog mailing list archives

Re: Dual stack IPv6 for IPv4 depletion


From: Owen DeLong <owen () delong com>
Date: Fri, 10 Jul 2015 09:23:36 -0700


On Jul 10, 2015, at 00:59 , Joe Maimon <jmaimon () ttec com> wrote:

There has been tomes on this topic. There will continue to be many more.

That is because many of you continue in trying to defend the following concept.

customer subnet bits  ==  isp customers bits

So now, the ISP is supposed to put some effort and gain more bits. Why not the customer?

Its inherently suspicious. Because its inherently wrong - for the ISP, and possibly for the address space as well.

Indulge me as I wax poetic.

I venture to say that proponents want to see everyone else have the service of their own dreams. When broadband 
rolled to the masses with a single ipv4 address per subscriber, forget about routing, their hearts broke. The new 
common denominator was a far cry from what their experience was. The division of internet into different classes of 
netizens a bitter pill to swallow. You are only one budget cut away from joining the ho-poloi. Its quite scary.

Hence the determination that no user should ever have to go without enough addresses ever again. A new common 
denominator, now is the time to get it accepted!

It will be like the old days, a class C with every leased line! Forever!

I will concur with most of this.


And the ISPs?

They have enough to get started and they can get more if they put the effort in.

Actually, as has been pointed out earlier by me and Valdis at least, they can get enough to last a good long time up 
front if they just do a little bit of napkin math before submitting their request.

Here’s how it works:

JimBob’s ISP and Bait shop serves their customers from 25 distinct wiring centers. They expect to deploy another 50 
wiring centers over the next 5 years.

Their largest wiring center supports 5,000,000 customers.

Representing 5,000,000 requires 23 bits. Rounded up to a multiple of 4 becomes 24 bits. At 24 bits, 5,000,000 is < 75% 
of the available /48s. So 24 bits is a good number.

Representing 75 wiring centers requires another 6 bits. Rounded up to a multiple of 4 becomes 8 bits. At 8 bits, 75 is 
< 75% of the 256 available numbers, so 8 is a good number.

24 + 8 is 32.

48 - 32 is 16.

JimBob’s ISP can apply to ARIN for a /16.

Other RIRs are a little different, but still usually not terribly hard to get a large allocation if it can be even 
remotely justified.


So all the rational and logical debate is pointless. Gut feelings, philosophy and emotions are what is at stake and 
those tend not to respond well to things like logic and reason.

Perhaps. Unfortunately, I think it is more the long-prefix crowd that is going on gut feelings. Unless you can show me 
how there’s harm to the ISPs from /48s per end site, or any other logic to support the need to retain the concept of 
second-class netizens, then I think logic is on the side of a more egalitarian internet.

Owen


Current thread: