On 2016/01/30 3:22 PM, E.Pasma wrote: > The diagram got broken in my email and here is another try: > > Needs to be light | Needs to be | Needs to do | > (small footprint) | Human-Readable | calculations | > ----------------- | ---------------| ------------ | > YES | YES | NO | Integer as > | | | Igor's suggestion > | | | > YES | NO | YES | Float/Int > | | | Julianday > | | | > NO | YES | YES | Datetime/Numeric > | | | ISO Standard
Thank you for the fix. > > With respect to Igor's suggestion, yyyymmdd (as integer), why not leave out > the century? I prefer the oldfashoned yymmdd. When dealing with a localized context around the current period, a two digit date is often enough - so if you see '12 or '16 or '20 you can easily assume that to mean 2012, 2016 etc. But what if you see '51? Would that be 1951 or 2051? The context would probably enlighten the meaning, but it's best to leave context to the users and not pre-empt it during the design phase. A four-digit year is best for standard human reference. (You won't need to also add the AD. bit) :)