On May 13, 2020, at 12:36 , Elad Cohen <elad@netstyle.io> wrote:

----

Do you realise that this means you're requiring changing *every*
socket-speaking application in the world?

----

Every internet host that will want to speak IPv4+ , will have an update (for example through the operating systems automatic updates mechanisms)

Uh… You left out Every Application on Every Host.

How do all of those applications get updated (some fo which are in-house custom code whose maintainers long since retired).

If you’re recoding all the applications, that’s pretty much the last hurdle left at this point for IPv6, so what’s the advantage of IPv4+ at this point? None whatsoever.

----

It's taken us decades to get applications to use the new struct to support
IPv6+IPv4, resetting the timer back to 0 and starting over does not help
deployment. It just kicks it another 20 years down the line.

----

I wrote about the usage of a roundtable in order to implement everything fast (the roundtable will include one representative from each of the operating system vendors, one representative from each of the routing equipment manufacturers and one representative from each of the 5 RIR's), if I will be elected to RIPE board I will do everything in my power so this roundtable will be formed fast and that the needed updates will be created fast. Each party in the roundtable will receive an amount of free IPv4 addresses from the new IPv4+ pool, and each ASN will also receive for example a /21 , home-routers and home-modems will not be needed to be updated and they will support IPv4+.

----

You're just inventing yet another incompatible standard and you have to
touch everything, DHCP, DNS all applications etc.

----

There is an adjustment to IPv4+ that the format of addresses will not be [0-655365].[0-655365]v4 - but it will be [256-511].[0-255].[0-255].[0-255]

So IPv4+ addresses will be in the format of IPv4 addresses - it will end-user adoption of IPv4+ easier and also integration in the applicative layer easier (as application developers will only need to set that the first number can be higher instead to support a new format of [0-655365].[0-655365]v4 )

So you want to retain all the need for NAT, continue to make end-users second-class citizens, avoid most of the gains and improvements of IPv6, while reducing the workload involved in adoption by (oh, wait, you’re increasing the workload required)…

Where’s the win in this, exactly?

Owen

From: Mikael Abrahamsson <swmike@swm.pp.se>
Sent: Wednesday, May 13, 2020 10:22 PM
To: Elad Cohen <elad@netstyle.io>
Cc: Brielle <bruns@2mbit.com>; NANOG list <nanog@nanog.org>
Subject: Re: RIPE NCC Executive Board election

On Wed, 13 May 2020, Elad Cohen wrote:

> LOL funny seeing you changing your mind by 180 degrees when someone you
> know in the community writing to you the exact same thing.

"In addition, the sockets API should be extended to support IPxl with a
new socket domain PF_IPXL which is identical to PF_INET in every respect
save that the IP addresses are 8 bytes long instead of 4."

Do you realise that this means you're requiring changing *every*
socket-speaking application in the world?

It's taken us decades to get applications to use the new struct to support
IPv6+IPv4, resetting the timer back to 0 and starting over does not help
deployment. It just kicks it another 20 years down the line.

You're just inventing yet another incompatible standard and you have to
touch everything, DHCP, DNS all applications etc.

--
Mikael Abrahamsson email: swmike@swm.pp.se