jbpros · Oct 18, 2026 · Oct 17, 2026 · Oct 18, 2026 · Oct 18, 2026
diff --git a/REFERENCE.md b/REFERENCE.md
diff --git a/ergo.rego b/ergo.rego
diff --git a/ergo_test.rego b/ergo_test.rego
diff --git a/spec/OPEN.md b/spec/OPEN.md
@@ -44,6 +44,13 @@ Derived values on a side of `compare` (`count`, `sum`, `add`, `literal`, and `wi
 
 A subject with `of` (see [REFERENCE.md](../REFERENCE.md#a-subject-built-on-another)) has an entry in `policy/schema.json`, but no rules and no cases yet.
 
+## Parts
+
+A subject's parts (`has`, see [syntax.md](syntax.md#parts)) leave a few things open:
+
+- `named by` a bare word that is a part's alias is written wrong. It would have to read the part's value and use it as a key, as it does for a `for` alias, and nothing says yet whether a part that is a path with its own `named by` can be used that way.
+- `identified_by` of a part is read and shown, but nothing uses it. A later change is meant to tag the values a check reads with the id of the part they sit in, so a report can say which one.
+
 ## Time limits
 
 `time.format` only takes years 1678 to 2261 and only counts the first nine digits of a fraction. Both limits come from OPA, which holds a time as nanoseconds since 1970 in a 64-bit integer. A port in another language would have to add them on purpose. Either they stay, and the spec keeps them as a rule every implementation follows, or the spec drops them and `ergo.rego` has to compare times some other way outside that range.