Community Wishlist Survey 2017/Wikidata/Stop making false claims about dates

Stop making false claims about dates

Discussion edit

  • It is currently impossible for an editor to enter an accurate birth date for anyone who wasn't born where the offset from UT was 0 at the place of birth (the UK in winter, for example). It is beyond the ability of editors to fix this. The developers should allow dates in local time to be stored.
Let me give an example. I read in a reliable source that a certain person was born in Australia on July 15, 1975. The date of birth is given, but not the time of day. If the person was born at 1 AM local time, the UT birth date is July 15. But if the person were born at 11 PM local time, the UT birth date is July 16. The editor can't solve this, it's on the developers. Jc3s5h (talk) 13:09, 16 November 2017 (UTC) edited for clarity 15:11 18 November 2017 (UT)[reply]
  • This is a symptom of a more fundamental issue: it's not possible to store full datetime values in Wikidata, you can only store the date and it assumes that the time is 0h UTC. It would be much better to allow the full datetime to be edited, so that the hour/minute/second can be specified as well as the day. Thanks. Mike Peel (talk) 20:47, 18 November 2017 (UTC)[reply]
  • For most time properties we have, like date of birth or death, you are not going to find sources with dates more precise than a day. Some obituaries might have time of death but it would be provides in local time, with whatever time savings adjustment would be proper for the place and era. So a lot of dates even if you know the time would be hard to convert to Universal Time. --Jarekt (talk) 12:56, 20 November 2017 (UTC)[reply]
  • With some rare exceptions, all dates are timezone-less. We export them as dateTime, but in fact virtually all of them are just dates, with no time. Since base JSON model still pretends they are date-times so does RDF model, but maybe it's time to refine and change that. What I am absolutely opposed to and think is the worst thing we could ever do is start converting dates to "local timezone" and "UTC". That would lead to utter insanity - we do not have historical data on timezones, and even if we had and somehow managed to make all data conversions work (which we with 99.9999% probability can't) it would result in converting all our data to utter junk as nobody ever cared what was the date in Greenwich, UK at the time certain event happened - unless that event happened in Greenwich, UK. What everybody cares about is what the date was in the place where event happened, and that's the only thing we should ever deal with. There should absolutely be no such thing as "UT birth date". What we have now is we deal with dates right, but we record them wrong - we pretend as if they are date-times in UTC, where they are just dates, without time. That is something that we may want to change - and it probably requires wider discussion in the community. It's certainly not a developer question until we decide to remove the pretense of having "time" part from dates - at which point yes, the developers should take note.
However, this may make date-times and dates (if we ever have proper date-times) incomparable - which the effectively are, absent historic timezone data since beginning of the universe, but may be inconvenient for practical reasons. Or, we decide we give up on times altogether (given that we didn't use them for years now and still are fine) and eliminate times from the data model.
In any case, I support starting a proper discussion on this. Laboramus (talk) 21:05, 28 November 2017 (UTC)[reply]

Voting edit