Interface iteration

[James Browning]

On Wed, Jan 13, 2021, 4:21 AM Hal Murray via devel <devel at ntpsec.org> wrote:

> While looking for old mail tangled with #690, I stumbled into this, from
> Eric.
>  Did we ever do anything with this?  This seems like a wonderful
> opportunity.
>
> The catch may be that ntpd needs to know which interface a packet arrived
> on,
> actually the destination IP address.
>
> ----------
>
> You missed some interesting discussion on IRC last night (for after
> release).  While working on the CVEs Daniel has spotted that all the
> iteration-over-interfaces stuff (which is 75% of ntp_io.c and almost
> all of what's left of the ISC library) is unnecessary if we trust
> IPv4/IPv6 wildcard addresses to work (which are POSIX).
>
> This is kind of a big deal not just because of the LOC reduction but
> because it would let us sever a major and several minor platform
> dependencies - not just getifaddrs but the routing-sockets stuff as
> well.
>

My rat brain is whispering 'no'. Something about link-local addresses,
people having retrograde configs, broadcast, and multicast. Two out of four
is not too bad.

I think this is the third(?) Time this has come up. The previous time seems
to have been May/June 2018. [1][2]

[1] https://lists.ntpsec.org/pipermail/devel/2018-May/thread.html
[2] https://lists.ntpsec.org/pipermail/devel/2018-June/thread.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ntpsec.org/pipermail/devel/attachments/20210113/0e60392a/attachment.htm>