When it was stamped.
There are two times on every event, and the difference between them is the whole point. The event time is when the handoff happened, as the sender reports it. The record time is when the gateway stamped it, and that one the sender cannot set.
Event time, and record time.
The sender says when the pickup happened. The gateway says when it received the record of it. Both are kept, distinctly, because a gap between them is itself evidence: a scan reported a day late is a different fact from one reported live, and a custody argument turns on which you are looking at.
Stamped at the door, not by the sender.
recordTime is applied at the gateway. Whatever the sender wrote in that field is stripped: the time it was actually recorded is a property of the door, not a value a party gets to assert. This is a law of the spine, not a setting, and like every law it cannot be renegotiated later.
The standard gives you this dimension. Here is what it still can't tell you.
EPCIS 2.0 §7.2.2 defines five event dimensions — and no performer is among them. The record can say a company did a process. It cannot say who observed the event. That answer has its own door.
when: answered — eventTime 06:12:04Z · recordTime 06:12:05Z who: (no field exists in the standard)
See your first trace.
Your capture workspace is provisioned from the seat list — we write when your seat is ready.