Documentation of eventDate

[wilsotc]

One of the interval examples for eventDate is inconsistent with ISO 8601-1:2019. The date interval includes the first instant of the starting unit and excludes the first instant of the end unit. Therefore 1900/1909 would include the first instant of 1900 and exclude the first instant of 1909 as a 9 year interval. This is also consistent with the duration being derived by subtracting the end unit from the start unit yielding in this example 9.

The decade of 1900 would be the interval 1900/1910.

The line:
1900/1909 (some time during the interval between the beginning of the year 1900 and the end of the year 1909)

Should be:
1900/1909 (some time during the interval between the beginning of the year 1900 and the start of the year 1909 exclusive)

[tucotuco]

Good eye, @wilsotc. Excellent. Thank you for this erratum.

[wilsotc]

I'm completely sure that this should be the way this is interpreted however the ISO document I have is very ambiguous on this point and even says that the start and end point in time should both be included! That does make software used to interpret it very difficult, and includes the possibility of a point in time being included in two adjacent intervals.

[wilsotc]

I suspect that while not ideal, the documentation may remain as is. ISO 8601-2 has a less ambiguous way to define the example: [1900]/(1909], but this would be unnecessary if we could interpret the interval as inclusive of the start and exclusive of the end point.

[MattBlissett]

I hadn't realised the ambiguity when I implemented this for GBIF.org's API/website, probably because I read the examples in DWC, the data quality task group, and maybe Wikipedia. Are you reading the actual standard document? I haven't purchased it.

In GBIF's use of eventDate, I understood 1900/1909 to be from the start of 1900 to the end of 1909, though with less precision than writing 1900-01-01/1909-12-31 or 1900-01-01T00:00:00/1909-12-31T23:59:59.

The BDQ group has the same:

https://github.com/FilteredPush/event_date_qc/blob/ebb24b8d6c187eddcbba0eef1ae60ba2e0c7202e/src/test/java/org/filteredpush/qc/date/DateUtilsTest.java#L213-L227

Would you also understand 1900-01-01/1909-12-31 to mean 1900-01-01T00:00:00/1909-12-31T00:00:00?

If there's ambiguity in the standard I'd prefer to keep the start-of/end-of meaning, as it fits better with what a user understands when they see "1900–1909" on a box of specimens (latest possible collection date of 31 December 1909).

[tucotuco]

Unfortunately, it is explicitly stated in Clause 5.5.3.2 "Time intervals — Representation by start and end" of ISO 8601-1:2019:

"The start of the interval is included, and the end of the interval is excluded."

Wikipedia is brief, but also clear on the subject:

"A time interval is the intervening time between two time points."

With further evidence in the pattern /, so 1900/1909 includes the interval before 1909 begins, starting at the first instant of 1900.

The commentary around the example should say:

"(some time during the interval including the beginning of the year 1900 and the before the beginning of the year 1909)"

[chicoreus]

In order to represent "some time during the interval" rather than "the interval", don't we need to use the EDTF extension? Under EDTF, the semantics of 1900/1909 would be: a time interval with calendar year precision, beginning sometime in 1900 and ending sometime in 1909.

[chicoreus]

@tucotuco Does 5.5.4, "representations other than complete" provide more guidance?