Re: ifdef inet

[Maxime Villard <max%m00nbsd.net@localhost>]

Le 19/04/2018 à 19:22, Greg Troxel a écrit :
> Roy Marples <roy%marples.name@localhost> writes:
>
> > On 19/04/2018 12:34, Maxime Villard wrote:
> > > It would be good to decide whether we should keep the numerous #ifdef INETs
> > > in our code. I've seen many files that inconsistently use #ifdef INET; one
> > > place has them, another doesn't, and so on.
> >
> > I think we should aim to keep it and even make it work.
> >
> > Maybe one day INET6 will become prevalent enough to drop INET, or npf
> > will grow NAT64 so a pure INET6 network can talk to an INET one.
>
> I agree we should keep it.   Today, it's reasonable to want to build a
> kernel with INET6 but not INET, for testing, to be hard-core v6, to
> build a v6-only firewall, and probably more that I haven't thought of.
>
> It's also reasonable to want to build a kernel with no networking.

Alright. I think we are far, far from that - IPv6 heavily relies on IPv4 -,
but in the end there are good reasons for wanting an IPv6-only firewall. So
I'll leave the #ifdefs as they are.

A ~similar, but not totally related, topic: it would be good to modularize
our "external" protocols, like IPsec, CARP, PFsync, MPLS, SCTP and so on.

Maxime