stzServiceRegistry

(sans commentaire dans la source : le nom se lit comme une phrase) initaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) nameaucune extension

the dependency surface ------------------------------------------- Declare a service this solution depends on, before anything is bound. This is what turns "what does this touch?" from archaeology into a query -- and it is what lets an UNBOUND dependency be a finding rather than a surprise.QCSXTZZZIBWUMany

(sans commentaire dans la source : le nom se lit comme une phrase) is declaredaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) declared servicesaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) number of declaredaucune extension

binding ---------------------------------------------------------- Bind an implementation. The posture is asked OF THE OBJECT: anything with IsSandbox() answering true is a sandbox. Declaring itself beats guessing from a class name, and it means a third-party double can opt in.QCSXTZZZIBWU

...say it explicitly when the object cannot.QCSXTZZZIBWU

Bind a genuine LOCAL equivalent -- sqlite, the filesystem, a local model. Not a fake, so unlike a sandbox this may ship; see the posture note above.QCSXTZZZIBWU

Bind the real thing, naming the STORE SECRET its credential lives in. The name, not the key: a registry that held credentials would be one more place they leak from.QCSXTZZZIBWU

Remove the IMPLEMENTATION but keep the dependency. The service is then declared-and-unbound, which IS a finding -- your solution still needs the thing, it just has nothing to serve it. To retire the dependency itself, use Undeclare.aucune extension

Retire the dependency altogether -- the solution no longer needs this service. Unbinds it too, so nothing is left half-declared.aucune extension

resolution (what the application actually calls) ------------------ The service by name. RAISES when nothing is bound -- an unbound dependency that silently returned NULL would fail later, somewhere else, as a null-call with no hint of the real cause.aucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) hasaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) bound servicesaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) number of boundaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) posture ofaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) is sandboxedaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) secret name ofaucune extension

every service still bound to a fake -- the "what is not real yet" list.aucune extension

the genuinely-local ones: real, self-hosted, shippable.aucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) is localaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) live servicesaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) unbound servicesaucune extension

the phase -------------------------------------------------------- :development / :emulated -> sandboxes are expected. :production -> they are a violation. The phase is the SAME switch stzDelivery already uses, so nothing new decides what is real.QCSXTZZZIBWU

(sans commentaire dans la source : le nom se lit comme une phrase) phaseaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) is productionaucune extension

governance ------------------------------------------------------- [ [ :invariant, :severity, :where, :message ], ... ] -- the same shape stzSecurityPosture and the graph rules use, so one CI gate covers all three. poStore may be NULL when there is no keyring to check against. NOTE: each _Check returns its OWN list and this LOOPS to append the elements. Haro's `+` on a list appends the whole list as ONE element, so `_aF_ + This._Check...()` would nest rather than accumulate -- and a nested findings list silently reports zero errors. Same shape as stzSecurityPosture.aucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) findingsaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) is sound viaaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) is soundaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) number of findingsaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) report viaaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) reportaucune extension

ASKED IN THE PRODUCTION FRAME, whatever the current phase. "May I go LIVE?" IS the production question, so answering it from a :development phase would have said YES with a fake still bound -- the phase-dependent invariants (sandbox-in-production, ephemeral-in-production) simply had not fired yet. The phase is set, the surface judged, the phase restored: asking is not declaring. (Found while writing the narration; the guard had only ever asked after setting production, so the honest answer and the convenient one agreed.)aucune extension

the surface judged as production would judge it, then put backaucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) findings for productionaucune extension

...and the same, but explaining itself.aucune extension

NOTE this does NOT duplicate the invariants above. Findings from Findings() already carry the report's shape, so the registry joins the shared CI gate directly: oReport.IngestLegacy(oReg.Findings(), "services"). What the GRAPH adds is the question a flag check cannot ask -- which PART of a solution depends on a fake -- and that needs the delivery's parts joined to these nodes (see stzDelivery.AsRuleGraph and stzServiceRuleSet).aucune extension

(sans commentaire dans la source : le nom se lit comme une phrase) showaucune extension