A Farewell to ARPs: IPv4 Service on IPv6-Only Networks

labs.ripe.net

55 points·speckx··8 comments

ARP is not an issue at smaller scales, but at larger ones it can cause problems. Reducing the L2 broadcast 'blast radius' is one of the touted benefits of EVPN (which use may be worth considering once you get above ~20 switches):

* https://www.youtube.com/watch?v=W_z37Rq7qgY

Saw this in IETF already but it's not clear why it's useful especially as the old path still needs to be implemented for compatibility. TFA is obviously LLMslop, which is very disappointing from RIPE.

I don't get your question? If you use a /32 then there is no broadcast address and no ARP. And the /32 are then announced by your internal routing protocol.

For corporate networks, where you control what devices get onto non-guest VLANs/subnets, you'll have better control over devices that support this new functionality. See for example Google's work on removing IPv4 on corporate networks:

* https://www.youtube.com/watch?v=UTRsi6mbAWM

Even in organizations where IT cannot control devices, like universities which are largely student-BYOD, IPv6 take-up can reach 80%:

* https://www.youtube.com/watch?v=2B-liebzcOMmkm&t=10m

"Obviously" in what way?

But why is this little hack advantageous over dual-stack? Advanced switches already don't forward ARPs they can handle locally.

Because dual-stack means having to deal with stand up IPv4, which entails more configuration, more address planning, more infra, etc. See perhaps "§5.1 IPv6-only Compared to Dual-Stack":

* https://datatracker.ietf.org/doc/html/draft-ietf-v6ops-6mops

Fewer IPv4 deployed subnets means those addresses can be used where they're 'really' needed instead of being 'wasted'.

> That's real, but it understates the problem.

It's all so tiresome.

What drives people to post blatant slop like this?

Reply on news.ycombinator.com