Questions

When several State Restrictions in one partition are satisfied at once, which State is the object in? The blog's own Artist example is satisfied by every later state, so state() returns a set, not a value. Is the intent (a) the most specific satisfied state, (b) mutually exclusive restrictions enforced at authoring time, or (c) something else?

Does a Thing have exactly one state dimension, or several independent ones? Claim 12 says "quasi-orthogonal object state machines", which suggests several. If one, how are genuinely independent lifecycles (payment status vs fulfilment status) modelled?

May a State Restriction reference another object's state? If so, are mutually recursive state definitions intended, or must the dependency graph be acyclic?

Is state membership asserted into the graph (rdf:type), or purely derived? Assertion is monotonic and never retracts, which conflicts with lifecycle progression.

Is state evaluated at business time, at transaction time, or both? The blog says "at a particular point in time" without saying which clock.

Is the transitioning instance always the Object of the SPOP triple, per claim 1? Can a Filter gate the Subject side without an inverse property?

Are an Operation's Filters conjunctive, with disjunction expressed as multiple Operations?

In a composed Filter, which model declares the predicate — the Subject model, the Object model, or the parent?

On import, the child's Main Thing becomes a subclass of the parent's Main Thing. Does the child then inherit the parent's Properties, States and Operations? Inheriting States would multiply partitions.

Can one Operation transition more than one object? Claim 4's "storing optionally the second object data" suggests yes; if so, what is the atomicity?

Is Authorization checked before or after Filters? Checking after can leak other objects' states through error messages.

Derivation bundles five unrelated things (admin parameters, value sets, formulas, scoped data sets, external objects). Was that deliberate, or would you accept splitting it into five constructs?

Can two Derivations in one Operation write the same property? Resolution order, or an error?

What happens to live instances when a release changes a State Restriction and they silently reclassify — possibly mid-lifecycle into a state with no valid exit?

Multi-tenancy: is a Universe per tenant, or an industry-shared asset that tenants instantiate?

Are Outbound Events guaranteed delivered, ordered, and to whom — Agents only, or other objects?

Can the same Object model be imported twice into one Universe with different Filter extensions? If so, what is the identity of the resulting Things?

Bridges are described as "an ontology dedicated to expressing the mapping". What makes a bridge correct? Is there an obligation it must satisfy?

Does a Component Filter extension that makes a previously-reachable Object state unreachable count as legal, a warning, or an error?

Do you still have the glossary content? The plugin is dead, the index page was already empty in 2013, and only 7 of 17 term definitions survive in the Wayback Machine. Your own controlled vocabulary is the highest-value missing artefact.

Two documents exist in the archive but not on the live site: /business-capability-definition-using-pciml/ and /reference-reading-guide/. Were they withdrawn deliberately?

The Method page defers the Technical API and Platform API to "see Post (to be published)". Were those ever written?

Eight images named Why-PMDA1-7 are on the site but referenced by no page. Is that an unpublished deck? Is there a fuller version?

The live site reports zero comments, but archived ?replytocom=187 and 188 show comments existed on The Product. Were they deleted, and is there anything in them worth keeping?

Is there unreleased material — code, the prototype, tightened semantics, notes? Is any of it subject to an obligation to someone else?

PAT:us9378071b2#para-2014 fixes the Filter's quantifier as existential in the engine. Does the notation allow a modeller to ask for anything else — EVERY, NONE, a count comparison — or is existential the only proposition PCIML can state?

pciml:hasPrerequisiteSubject and pciml:hasPrerequisiteObject induce a partial order over a Thing's States. Where several State Restrictions hold at once, is the intended reading "the maximal State under that order"? And is the order guaranteed acyclic?

The worked Filter table transitions the proposition's Subject in one row (Activate Artist) and an object in neither position in two more (the Instantiate rows). So what determines which object an Operation transitions — is it named on the Operation, or derived from the proposition by a rule we have not found?

Both worked Derivation formulas are ambient calls — afn:now() and pmda:User(). Was that intended as the general case for Derivation formulas, or a convenience for the two audit fields? If general, the method cannot support bit-identical replay without a change.

The Condition table has a Condition Type column with values Error and Warning. Does a Warning Condition that does not hold still permit the transition? If so there are two Condition semantics and the spec records only one.

PCIML has one Property construct doing two jobs: data slots that State Restrictions read, and relationship predicates that Filters traverse. Was that deliberate, or is the distinction implicit? A tool almost certainly has to separate them.

Are Object, Component and Universe a normative three-level hierarchy, or just the example depth of a relation that recurses without limit? The single-mechanism claim suggests the latter.

Are viewpoint, scope and abstraction the complete classification scheme, or is there a fourth axis — which capability a model belongs to? Widening this later means migrating every model in existence.

The description names a PCIML construct Resource (801) once, at [para-1954], as "the construct Resource 801 that describes the Object", and never again. Is it an earlier name for Thing, a superclass of Thing and Property, or something dropped before filing?

The Condition table's Enforces Property in Condition column holds property names in rows r3-r9 and State names in rows r10-r14 (New Artefact, Created Artefact). Is the column used inconsistently, or does it mean something different for Artefact rows? As transcribed, every Artefact Condition is unevaluable: there is no property by that name to require.

Which State does a given Operation achieve? pciml:achieves is named in the patent, but none of the three transcribed blog tables records the Operation-to-State pairing — the Filter table names Operations, the Condition and Derivation tables name target States, and nothing joins them. The demo currently infers the pairing from naming convention and says so on screen. Is there a source that gives it, or was it only ever implicit?

The Condition table gives the State as Create Artist in rows r4-r6 (an Operation name; every other c3 cell holds a State) and as CreatedUser in rows r8-r9 (one word here, two in c1 of the same rows). Neither is the name of any State. Were Created Artist and Created User meant? As transcribed, five of the twelve Conditions attach to no State and can never fire, and the demo reports them as unattached rather than correcting the spelling.

Fix the ontology IRI base. It must resolve until at least 2033 and is irreversible. Candidates: w3id.org/pmda (community-maintained, free, redirect controlled by a pull request) or a domain owned indefinitely. Never a hostname that might lapse. Must be settled as ADR-0002 before a single triple is written.

Set up the two off-repo copies of record: an object-storage bucket with versioning and object lock, and an offline drive. A single repository is a single point of failure for the evidence.

Re-examine any assertion that cited snapshot 2026-09-08T093017Z before it was found to be truncated. Currently none exist, so this is cheap now and expensive later.

Submit every source URL to the Wayback Machine and archive.today and record the returned URLs. wayback_url is currently null throughout citekeys.json, which is a recorded gap in the citation scheme's third durability layer.

Ten of the seventeen archived glossary term URLs are still unrecovered because Wayback rate-limits. Re-run tools/corpus/archive.py until the set closes, or record which are genuinely unarchived.

Decide whether transcription of the 53 patent drawing sheets is worth its cost before the semantics are settled, or whether the figures should wait until a specific assertion needs one.

What is a Filter's quantifier? "Created Artefact isAuthoredBy Active Artist" has four incompatible readings: SOME, EVERY, NONE, COUNT. No source states one.

Is a Condition evaluated against the data before the Derivations run, or against the proposed post-image? Claim 2 puts validation inside the transition calculation; claim 4 puts storage after the Condition is true.

When a Component "replaces the behaviour required" of its constituent Objects, is that override (replacement) or refinement (strengthening only)?

Is there any time-triggered transition ("after 30 days, expire")? The construct set has no timer. If yes, how does it stay deterministic?

The AU specification's description is materially shorter than the US one (47k vs 67k words). Are they the same disclosure?

Is the AU description the same disclosure as the US one, or an artefact of how Google renders AU documents? Must be settled before any assertion cites the AU description.