TIME: Suggestions for alignment with PROV-CONSTRAINTS "precedes"?

[ajnelson-nist]

This is an Issue about how to interpret a certain property in OWL-Time, time:before. There is a detour through PROV that explains why I'm asking. It is possible this issue also shows a point of needed clarification in the current OWL-Time draft.

While reviewing the non-normative alignment of TIME and PROV here ...

https://github.com/w3c/sdw/blob/gh-pages/time/rdf/time-prov.ttl

... I saw that there is an alignment axiom for prov:Activity:

prov:Activity
  rdfs:subClassOf time:TemporalEntity ;
.

From working with PROV, I had come to think it would be more appropriate for prov:Activity to be a subclass of time:ProperInterval. However, when double-checking my sources, I wasn't able to find a strict requirement that, given a prov:Activity with a prov:Start and prov:End, that prov:Start was required to happen strictly before prov:End.

From review of PROV-CONSTRAINTS, I came across this evidence that a prov:Activity is permitted to be instantaneous:

• Constraint 30 (start-precedes-end) uses the defined term "precedes".
• "precedes" is defined like so: "Specifically, precedes is a preorder between instantaneous events. A constraint of the form e1 precedes e2 means that e1 happened at the same time as or before e2." So, "precedes" is roughly writable as "<=".
• "strictly precedes", in the same paragraph as "precedes", is defined to have the two related (instantaneous) events not occur at the same time.
  • The only usage of "strictly precedes" that I could see in PROV-CONSTRAINTS was Constraint 42 (derivation-generation-generation-ordering).
• Reviewing "Activity" in PROV-O and in PROV-DM didn't provide a hint on required inequality of its start or ends.
• Reviewing "Start" in PROV-O and in PROV-DM shows a reliance on "follows" (inverse of "precedes") that is not clarified as being a strict "<" or permissive "<=" binary relationship.

Where all this circles back to OWL-Time: I see two properties that can be used to relate one time:Instant to another time:Instant, time:before and time:after. The definitions, as currently worded, do not make explicit whether the properties permit equality of the instants being related. Personally, my normal reading of the words "before" and "after" imply a strict-inequality relationship; but, apparently W3C editorial policy in the past has permitted a standard to be posted where "precedes" could be defined as "<=", so now I feel the need to check.

If I have a statement x time:before y ., can x and y refer to the same time:Instant?

For the sake of finding a way to align with the "precedes" definition in PROV-CONSTRAINTS: Is there a predicate in OWL-Time that defines "<=" between time:Instants?

Should one of time:before or time:after be designated an owl: AsymmetricProperty? An owl:ReflexiveProperty or owl:IrreflexiveProperty? time:before is currently a owl:TransitiveProperty, so it seems some thought was put into property specializations, but I'm not sure where notes on those specializations might be.

[dr-shorthair]

Thanks for the very detailed comment.

We only used some of the possible OWL axioms, in particular we did not attempt to complete the possible owl: AsymmetricProperty, owl:ReflexiveProperty, owl:IrreflexiveProperty, owl:FunctionalProperty, owl:InverseFunctionalProperty.
Maybe we should have.

However, the issues you raise may be dealt with in the 'extension' to OWL-Time where we attempted to complete the set of topological relations - see https://www.w3.org/TR/vocab-owl-time-rel/ in particular see https://www.w3.org/TR/vocab-owl-time-rel/#new-relations

As you will observe, this work, though several years old now, has not progressed past 'Note' status.
This was partly because we needed more evidence that these additional relations were in fact needed.
Your comment appears to confirm that they are.

[ajnelson-nist]

@dr-shorthair , thank you for the references. I checked them again, and think I found an answer on whether time:before is meant to be strict-< or the more-permissive <=. In short: It seems to mean strict-<.

I noticed time:disjoint from the Extensions Note used a could-be-clarifying phrase, boldfaced for emphasis:

If a temporal entity T1 is disjoint with another temporal entity T2, then the beginning of T1 is after the end of T2, or the end of T1 is before the beginning of T2, i.e. the two entities do not overlap or coincide in any way, but their ordering relationship is not known.

I appreciate that's a non-normative definition in a Note, but looking for just the words "not overlap" led me to time:intervalDisjoint in the current Recommendation, which includes "... the intervals do not overlap in any way". Checking on the definitions of time:before and the illustrations in Figure 2, it appears to me that time:before and time:after do indeed mean strict-inequality. Else, time:intervalDisjoint could be used on two proper intervals with an overlap of exactly their boundary instants.

So, it seems time:before and time:after would be worth typing as owl:IrreflexivePropertys as a point of clarification.

It also seems precedes from PROV-CONSTRAINTS is a property not currently represented in OWL-TIME or the Extensions Note, but could be done as a parent property of time:before and (the Note's) time:equals, with appropriate parallels for proper intervals.

[ajnelson-nist]

I think it would also be helpful to add time:after owl:propertyDisjointWith time:before . to help clarify the non-overlap of temporal regions implied by the two properties.