[arin-ppml] Solving the squatting problem

[Owen DeLong]

> On May 16, 2019, at 2:16 PM, Jimmy Hess <mysidia at gmail.com> wrote:
> 
> On Thu, May 16, 2019 at 1:52 PM Owen DeLong <owen at delong.com> wrote:
> 
> It doesn't really matter...  ALL of these software kernels receive
> updates frequently;
> mobile and desktop OSes in particular have numerous updates per month, and
> even BSD, Cisco, Juniper, Arista  OSes  have frequent updates being made.

It matters… Someone has to write the code. Someone has to test the code. Someone
Has to write the test plans. Someone has to field the bug reports. Someone has to
Fix the code. Someone has to write the test plans for the fixes…

There’s a whole lifecycle of support for making such a change that needs to be accounted
for and that takes resources. It’s my considered opinion (and I’m pretty sure there’s a
good number of people who agree with me that chasing the class E rabbit down the
rathole will be a waste of those resources which could be better spent focused on IPv6.

> Adjusting the disposition of 240/4   in the kernel is a minor change.
> Likely  less than  1% of the change volume  these systems' codebases
> receive during the average month.

Doesn’t matter.

> 1 or 2 lines of code for vendors to adjust --- not a huge deal
> (so long as it is software and not hardware/ASIC logic that needs to change).

I bet there’s not a single OS where this change can be made in so few lines of code.

Any UI that interacts with interface settings needs to be updated. Likely several
UIs that display IP addresses need updating. Any libraries that parse IP addresses
And any libraries that validate IP addresses and any code that performs a bounds
check on an IP needs to be updated. It’s nowhere near as trivial as you want to
claim it is.

> That is likely less than the amount of text that needs to be altered in the
> RFCs to state that  240/4  be reclassed as  global Unicast.

I bet that’s not actually true in the real world.

> Certainly not at a comparable level of complexity as implementing V6.

Implementing v6 is already done in every one of the platforms I mentioned, so,
I’d argue that configuring IPv6 on your network in parallel to your IPv4 infrastructure
is likely significantly less complex than deploying class E space.

YMMV.

Owen