Consistency on use of recurring date intervals

[philipashlock]

The use of date intervals was clarified with the use of ISO 8601 on #262 but that also needs to be updated in a few other places:

• the machine readable JSON Schema regular expression - https://github.com/project-open-data/project-open-data.github.io/blob/master/schema/1_0_final/single_entry.json#L183
• guidance on the relationship between this and accrualPeriodicity.

[dsmorgan77]

Right, and if the interval is <1 day, then accuralPeriodicity must be "Continuously Updated"

[dsmorgan77]

Someone just approached me with another exciting potential issue on the temporal field. I don't know why we would do this, but hear me out.

Suppose there's a dataset that contains two years of historical data. We update it monthly. In that update, we drop the oldest month and add the newest month. So, the dataset has a rolling start date.

To represent this in the ISO8601 format, would we simply represent a duration of P-2Y (always two years? This one makes my head hurt.

[dsmorgan77]

Another question on this. For a dataset that is completely irregular, but remains actively managed, does it really make sense to populate an end date? But there's also no way to write out an interval that's completely irregular in ISO8601. From a practical standpoint, the dataset's valid up through the current date, there's just no new record since the last update.

[philipashlock]

As part of the v1.1 schema update, the date fields have been normalized around 8601 and updated in the regular expression