On Mo, Aug 15, 2016 at 08:40:19 +0100, Sarah Hoffmann wrote: > On Mon, Aug 15, 2016 at 09:10:46PM +0200, Martin Raifer wrote: > > > Martin mentioned a deviation of exactly 4 seconds > > > > Well, actually, the 4 seconds were only for the one example changeset. > > In other changesets, I saw different time differences (most were even > > smaller if I recall it correctly). > > The reason for these time differences is that the clocks on some of > the API frontend servers were out of sync. TomH has fixed the > problem in the meantime. > > Only the changeset timestamps were affected. The timestamps for the > objects should be okay as they are created on the backends[*]. So you > should rather trust those when in doubt. > > [*] At least when using diff uploads which every decent editor is > doing nowadays.
So basically the changeset timestamps are useless. :-( Question is: Do we fix it? If yes, then how? Is it worth it? Do we need those timestamps really? To avoid this in the future we'd have to set the changeset timestamps in the database instead of in the web servers. And how about old changesets, do we retroactively fix them? Jochen -- Jochen Topf [email protected] http://www.jochentopf.com/ +49-351-31778688 _______________________________________________ dev mailing list [email protected] https://lists.openstreetmap.org/listinfo/dev
