On Mon, 13 Nov 2017 14:21:52 +0000 Erik Kline <e...@google.com> wrote:
> Alternatively, if we really want to make all, default, and ifname > useful perhaps we need to investigate a tristate option (for currently > boolean values, at least). -1 could mean no preference, for example. I think this would make sense in general, but on the other hand it would be quite a big change, and Nicolas' patch has the advantages of being small, keeping the global flag functional, and restoring the previous default behaviour out of the box when 'accept_dad' is disabled by the user for a given interface. Besides, this still wouldn't solve the case where flags are >= 0 and conflicting -- there, it's still debatable whether we want a logical OR or a logical AND. In the end, I would prefer either Nicolas' patch, or to get rid of the global 'accept_dad' flag altogether. Having a flag which does absolutely nothing, which was the case before 35e015e1f577, doesn't sound correct by any means. -- Stefano