SaraAbbott0026 · 19 Feb 2026, 15:12 UTC
Give each signal one writer and agree the full exchange with both authors. A held request with a matching acknowledgement can survive a slow poll, but only if both ends agree when each owner clears its own output.
21 pointsI'd fix the ownership before making the pulse longer. Otherwise you've given two programs more time to disagree about the same bit.
9 pointsWhat tells the application this is a new job, Sara?
8 pointsA job label, but it is reused after restart. That needs an attempt identity agreed with the PLC author as well.
19 pointsInclude run and attempt identity in request and receipt. Our offline replay stopped mixing restarts once both authors carried them through.
15 pointsWould the proposed acknowledgement mean accepted for processing or inspection complete? Separate names in a document still need separate events in the program.
2 pointsWhat should the operator do when receipt has been confirmed but the inspection result has not arrived? That is a different wait from a request nobody has received.
5 pointsAnd preserve the history on reconnect. An empty local list can make the same retained completion look new again, even with a useful job label.
17 pointsCould the two authors walk through one delayed poll together before anyone changes the code? Seems a cheap way to expose which arrow each expects next.
17 pointsRebecca's walkthrough should include the housing still at the fixture when the application loses contact, so nobody writes a restart instruction that assumes a fresh empty job.
11 pointsI would bring the covering operator into that explanation too, once the authors agree it. Otherwise the job can have immaculate internal names while the screen still offers a button nobody knows how to interpret.
18 pointsDoes the timeout leave the original request identifiable, or will a retry quietly replace it?
10 pointsKeep it identifiable. If a late acknowledgement arrives after timeout, it must still belong to that original attempt. A new attempt should not inherit it because somebody has stopped waiting.
13 pointsOscar: acknowledgement means accepted for processing, separate completion result later. Bruno: timed-out attempt remains identified. Authors are walking through delayed poll, duplicate receipt and late acknowledgement together.
6 pointsDon't let duplicate receipt start another inspection either. Same receipt, same work, however many times the connection delivers it.
14 pointsWalkthrough found the application cleared the request before the PLC had observed acknowledgement. Revised draft gives each output one owner and specifies the return-to-idle order. It includes unmatched and interrupted cases.
9 pointsHas the operator-facing distinction reached the screen draft yet, Sara?
8 pointsYes. Waiting for receipt and awaiting inspection result are separate on the draft, with the agreed assistance route. Implementation and testing remain to do; this closes the conflicting ownership definition, not the whole fix.
6 pointsUseful. I would like to hear whether the covering operator can explain those waits without either author translating.
5 pointsThis discussion is closed to new replies after six months without activity. Last activity: .