> Don't bother. This helper is no go. I disagree on the 'no go' -- I do think we should have this helper. The problem is it should only be used in some limited circumstances - for example when processing packets coming in off the wire with real times in them... For example for something like measuring delay of ntp frames. This is of course testable but annoying (create a fake ntp frame with gettimeofday injected timestamp in it, run it through bpf, see what processing delay it reports). Lets not make bpf even harder to use then it already is...
