Re: SIP with NAT traversal and STUN using NPF

Martin Husemann <[email protected]> writes:

> On Mon, Jun 15, 2026 at 06:04:00PM -0400, Greg Troxel wrote:
>> So in the end, did the STUN and NAT work as I was describing, and this
>> was just having NAT set up correctly, instead of some odd not-right NAT?
>
> I still think NPF is missing the dynamic symmetric UDP mapping option,
> but it is not required for my use case (will investigate this closer,
> I may be totaly wrong on this, or if not it would be a good feature
> to add, but needs better/other testing).

No argument from me.  NAT complexity knows no bounds.  Not sure how much
any odd scheme matters though.

> I am not completely sure what caused the strange effects I saw (and
> partly still see). Some of them seem timing or sequence related (old
> state both in NPF and the VoIP device persisting accross changes).
>
> But now I know I can get it working with plain NAT and will document
> the overall setup once it is polished.
>
> We could use more examples in /usr/share/examples/npf/, but this setup
> requires other hacks (additional dhcpcd hooks interacting with NPF,
> unrelated to the SIP/VoIP part), so maybe a wiki page describing the
> overall setup would be usefull. Or extend the NetBSD guide's NPF
> chapter.

Interesting, and yes, examples, guide, or wiki would be useful.