Re: ISO8601 for dates?

Hiyas,

Yes, though it would be good to formalize how to refer to years before 0
CE, including years with more than 4 digits. For example, 66 million years
ago is the end of the Cretaceous period. It's allowed in ISO8601 to have
more than 4 years (at least according to wikipedia) so long as there is
agreement in advance... we should establish that agreement for RDF  /
JSON-LD usage if possible.

Rob


On Tue, Oct 14, 2025 at 11:34 AM Benjamin Young <byoung@bigbluehat.com>
wrote:

> Hi all,
>
> There's an ancient issue in RDF land with modeling dates and times well.
> Generally, you have to "pick a lane":
> * do I want just a "date" (xsd:date) with no time expression
> * do I want a "date + time" (xsd:dateTime) where every date MUST come with
> a time
> * do I want just "time" (xsd:time)
>
> The pain shows up when you try to model/express/hold a date *and/or* time
> in a single field—which is not uncommon in event systems, etc.
>
> What most "just JSON" developers seem to want (or expect they can have) is
> something closer to ISO8601 where there's one date(and/or time) format "to
> rule them all" such that a single field can hold:
> * 2025-10-14 (date)
> * 14:04:51Z (time)
> * 2025-10-14T14:04:51Z (date and time)
>
> The Wikipedia page for ISO8601 has some more examples:
> https://en.wikipedia.org/wiki/ISO_8601
>
> I realize y'all know all of the above. ;) What I'm pondering is would
> JSON-LD/RDF benefit from a defined ISO8601 datatype to support these (and
> more)?
>
> There is, in fact, RFC5141 <https://www.rfc-editor.org/rfc/rfc5141.html>
> which defines URNs for ISO specs, so in theory, this could look something
> like the following:
> <#thing> <createdOn> "2025-10-14"^urn:iso:std:iso:8601
>
> Obviously, choosing this datatype does push the "understanding" of what's
> inside to...well...what's inside, so there are still valid reasons to use
> the XSD datatypes which provide a clearer "look before you leap"
> experience. However, what this ISO8601 datatype could provide is scenarios
> (which are all too common...) where `xsd:dateTime`is chosen but with the
> full expectation by the ontologists that the time may be set to `00:00:00`
> and *not* mean midnight, but instead mean "ignore the time".
>
> So, I realize this is an antique issue/topic and this may not be anything
> like a solution for all the problems...but I thought I'd propose it and get
> some feedback. :)
>
> Thanks for entertaining this idea!
> Benjamin
>
> --
> https://bigbluehat.com/
> https://linkedin.com/in/benjaminyoung
>
>
>


-- 
Rob Sanderson
Senior Director for Digital Cultural Heritage
Yale University