David Golden <x...@xdg.me> writes: > On Wed, Nov 7, 2012 at 3:44 PM, Andreas Koenig > <andreas.koenig.7os6v...@franz.ak.mind.de> wrote: >> >> But maybe I should tell you about another interesting observation I have >> not yet talked about: some cpantesters have wrong clocks. One tester's >> watch is 3 hours and three minutes off, another one's about one hour. I >> have seen several others only some minutes or some seconds wrong. >> Possibly somebody in the whole setup is relying on the times contained >> within the reports while others are relying on real time? > > Because some people run offline and then submit, the log could have > times in any order. Of course, people might just have clocks that are > off, too. > > We pretty much just accept whatever time submitters claim -- just like > we accept whatever *ELSE* submitters claim.
What about adding a date field for the timestamp when the report was received by metabase? Would it not provide a chance to sanitize the chain? Barbie could query for a field with guaranteed monotonously increasing values, and those reports with timestamps in the future could be silently corrected. -- andreas
