Unlike other relationships this has subclasses for the different types of derivation. But also a DerivationKind (though indirectly linked via DerivationType) with largely the same Library of instances (DerivedFrom, QuotedFrom, RevisionOf, SourcedFrom), though missing subclass DescendantOf (which is self-referentially defined).

BTW there's a missing word "one" in the following under Figure 24
"DerivationTypes support the definition of the expected kinds of derivations that might result in the generation of one EntityType from or more others."

It's unclear how to use Snapshot generally, and how the creationDate is supposed to tie in with the Occurrence(s) that updated it.

The scope of a snapshot is not clear, and EntityRelationship is not timestamped in the same way.
If an Occurrence creates and removes some EntityRelationships involving an Entity, how is that represented? ActivityOccurrence::resultingEntityRelationship is not sufficient.

It's not clear what isUnique means in practice. For example 8.4.9, what does it mean for a ResponsibilityRelationship to be unique?
I assume it means unique to a specific Entity (it should say) but, if isUnique=true, does it allow any time overlap of responsibilities? What if the responsibility is delegated - won't the original still hold?

It inherits source and target but does not explain the meaning of the direction. I'd guess that the target is the more canonical and the source is the alternate but it should be spelled out.

The definition says "The AttributedTo relationship is a kind of AssociatedWith relationship that captures the fact that an Entity was created or transformed by some unknown activity or action of a Party."

However even if the activity detail is known, there may still be value in giving someone specific (out of all those who may have played some role) credit or responsibility for the Entity e.g. in case of support or other inquiries.

In PPMN DerivedFrom::derivee is used to represent the Entity FROM which something was derived; however in English it means the derived element - which is the other named end.
Quoting a definition:
"In linguistics, a derivee is a word that is formed from another word by adding affixes or changing the shape of the root. For example, the word "runner" is a derivee of the word "run".

In general, the word "derivee" means something that is derived from something else. The specific meaning of the word will depend on the context in which it is used.

Here are some examples of how the word "derivee" is used in a sentence:

In mathematics, the derivative of a function is its derivee.
In linguistics, the word "runner" is a derivee of the word "run"."

These terms all inherit from SCE::SemanticElement which is supposed to have a conceptURI. Though not mandatory in the metamodel it seems bad practice that none of the official library elements have such.

If only as an exemplar. For example, the standard graphical formats, HTML, PDF, programming languages, XML, XMI etc

It adds nothing over EntityType

One says "Output indicates that the target Entity is an output of some kind of the Occurrence.."
and the other says "Product indicates that the source Occurrence produces or creates the target Entity during the course of the Occurrence."
It seems both Product and By-product should be subtypes of Output (should at least be documented if not supported by the metamodel) but there is no definition of what would make a Product.

It's unclear if either would encompass destroying the Entity.

That would allow for treating a whole chain as a single occurrence for purposes of abstraction/drill down, and address inconsistencies in the current metamodel (e.g. it would seem that a single Occurrence cannot have a subject Entity or interestedParty since they are properties specific to OccurrenceChain).

Many people will want to use PPMN just for tracking occurrences (which is the level that PROV works at). Understanding how to do this is made a lot more complex by the fact that all the metamodels are presented along with the type/process level information.
Proposal: have the instance/occurrence level information in their own self-contained metamodels and the Type extensions separate.

It's described in the text of 8.9 but not shown in a diagram or included in the metamodel. It seems useful, to represent the relationship from PROV.

This is one of the 4 classic types of process communication e.g. https://www.brcommunity.com/articles.php?id=b634 - we have the other 3 (input, output, enabler).

[IDEF0 uses different terms - Control instead of Guide and Mechanism instead of Enabler. Use of MVF would enable this]

At the moment people would be excluded, based on the definition "An individual concept or informational or physical artifact". However there may be use cases where it's required to track chain of custody of a person e.g. a prisoner, a baby/minor, a patient, a VIP, a witness, an asylum seeker, a dead body, a mummy. All being held and transported and transferred between parties.

It's not helped that the descriptions of the properties of OccurrenceRelationship use "dependent", and the properties of Occurrence (Table 14) use "dependency" for both predecessor (OccurrenceRelationships) and occurrenceDependency. It's pretty hard to determine what's canonical vs what's derived. I think that OccurrenceRelationship is intended to be derived from OccurrenceDependency (which involves the more detailed flow of Entities). Another confusing aspect is that OccurrenceDependency is used for outputs: the source of such Dependencies is the Occurrence rather than the Entity produced. Normally (e.g. in UML) the source of a Dependency would be an Entity and it's dependent on the Occurrence that generated/updated it. I think it might be better to rename it. Further confusing is that the PPMNLibrary has a PPMNRelationshipKind of Dependency.

At the moment it's self-referential "Custody indicates that the source element has custody of the target element."
There are nuances and legal aspects to consider.
For example, if someone is working for a company does the company or the employee have "custody" of the Entity they're working on? Legally it's the Company.
If an Entity is lost or stolen does the finder, or thief, have custody? Legally they do not, apparently.

It's not clear whether this should be in scope (probably not), but at minimum something should be said about it. There are other standards that could be used for contact info, e.g. vCard, and OMG Commons has basic support for identifiers, and the GLEIF ontology has registration information. In general there are several aspects not addressed, several important for provenance - especially being able to identify the party.

• naming
• registration (for companies)
• jurisdiction
• identification (many schemes, including for non-human agents)
• entity type (for companies, but could be used for minors/prisoners)
• date created/born and ceased
• software/hardware version (for non-human agents)

It repeats "is an input":
Input indicates that the target Entity is an input to the source Occurrence is an input during the course of the Occurrence

The definition "The time that the Occurrence had the dependency on the Entity” is unclear in usage when the Occurrences themselves have both a start and end.
I think the property could safely be deleted.

3 mentions in the text of 8.2 and 8.2.1: should be ActivityOccurrenceType

Occurrences may be linked to Locations regardless of a Party, so it does not make sense to include the Locations package in the Parties model.

Unlike PROV which makes it explicit.
All we have is creationDate, which could refer to a snapshot.

It's unclear what the sequence should be, or even the meaning of grouping the Occurrences into a Chain. Is the "succession" to be based on timing, dependency, causation? Is it an error if they are not in time sequence?
How are the occurrenceHistory related to occurrenceRelationships and/or Dependencies?

BTW the definition should say "in the life of some Entity" as opposed to "NamedElement"

Figure 52 should show AcquisitionKind inheriting from SemanticReference. the other vocabs do have this.

For example, I would have thought a Performer would always have an interest.

For best practice and consistency

The specification includes figure 11 Occurrence kinds but the diagram is not complete and is not needed. Document number is actually dtc/22-11-05.

SCE was removed from other BPM+ specifications including PPMN and established as a separate specification. PPMN needs to reference the machine readable files associated with the new SCE specification.

The file and namespaces associated with the Parties related machine readable files of the PPMN specification currently use PM in the name and namespace rather than Parties. This has caused confusion in the community. The name should be expanded to Parties to resolve the issue.

The BKPMN language is referenced in the spec, but it is not an official language yet