Closed simonff closed 5 years ago
ISO8601 was just a quick shot and I feared that there are strange values for periodicity out there. Not sure how we can handle that. Fall back to a simple string data type where you can put a description? That is the least desirable option though and we should try to find something better. How do other standards solve this? I think mostly with free text fields?!
Yeah, I think the special cases just need a special extension.
I don't want to do something other than ISO8601 for the core, and the special cases can just use it in a crappy way. And then they can have their own way of representing the more advanced things.
Issue is outdated as we removed dataset periods from the collection spec. In data cube we now have a "step" field for temporal data, which allows ISO8601 for regular spaced steps and null
for irregular steps. I'd say for other periods we need a custom field as Chris said.
In rare cases, a dataset has the period that cannot be represented in ISO8601. Two examples I've run into are CHIRPS pentads (http://chg-wiki.geog.ucsb.edu/wiki/CHIRPS_FAQ#How_does_CHIRPS_define_a_pentad.3F) and PDSI files (see https://www.northwestknowledge.net/metdata/data/early/2018/)
In both cases, there are multiple assets per month with duration 5 or 10 days, and the last asset covers the remainder - its duration varies from month to month.