Repository navigation
Replies: 2 comments
|
Thanks for the careful write-up and the patch — all four points check out against current 1. The reply timeout leaves an orphan admin — confirmed; the fix belongs in vta-sdk. Setup still passes 2. Transport choice — yes up front, but not as a fallback after the timeout. Falling back to REST after a TSP reply timeout can't work: by then the setup DID's one-time hand-off is spent, so the REST retry is refused. We'll let the operator pick the transport before connecting (passing it through as 3. REST drops the mediator — confirmed, and it's worse on 4. Mistyped VTA DID — confirmed, in vta-sdk. On a resolution failure the sdk guesses a REST URL from the DID string, so the real cause surfaces much later as "is this a VTA?". We'll make it report the resolution failure instead. We'll post the PRs here as they land and update this thread when the work is done. |
|
Both fixes have landed. Thanks again for the report. OpenVTC: #502
vta-sdk / vtc-service: OpenVTC/verifiable-trust-infrastructure#2219
As noted earlier, falling back to REST after a TSP reply timeout isn't offered, because by then the grant is spent. Choosing the transport up front is the supported route. Getting the fixes: the openvtc changes ship in the next openvtc release. The sdk changes reach If you still see it after upgrading, please open an issue with the setup diagnostics and the mediator log for the request id. |
Uh oh!
There was an error while loading. Please reload this page.
Per CONTRIBUTING.md, raising this here for triage before filing an issue. Is this a known limitation, and would you take an issue (or a PR) for any of the suggestions below? The mediator side is filed as affinidi/affinidi-tdk-rs#957.
Versions: openvtc 0.4.0 (also checked 0.6.0: same code), vta-sdk 0.62.1 (0.65.1: TSP, session and provision_client code unchanged). Mediator side: affinidi/affinidi-tdk-rs#957.
What happens
Against a self-hosted VTA that advertises TSP, DIDComm and REST,
openvtc setup(AdminRotated) always provisions over TSP.run_connection_testis called withforce_transport = Noneinstate_handler/setup_vta_actions.rs, and the preference is TSP > DIDComm > REST. When the mediator doesn't deliver the reply (affinidi/affinidi-tdk-rs#957), the wizard fails with:By that point the VTA has already created the long-term admin and used up the setup DID's one-time hand-off. The new admin's key exists only in the undelivered reply, so the VTA is left with an admin ACL entry nobody can use, and every retry needs a new grant and a manual
pnm acl delete.Suggestions
#vta-rest), aspnmeffectively can.dev-overrideswithOPENVTC_VTA_URL, and the runner's REST arm, which sendsConnected { mediator_did: None }) save the account with no mediator. A later join then falls back toLF_PUBLIC_MEDIATOR_DID(setup_sequence/config.rs,join_flow.rs).unexpected challenge response … (is this a VTA?). It should say that the DID didn't resolve.Workaround
A local patch to 0.4.0 that passes
Some(Protocol::Rest)and keeps the resolvedmediator_did:Result: on 2026-10-15 the patched build provisioned on the first try against the same VTA and mediator, about two hours after the TSP attempt failed. It authenticated over REST as the setup DID, the AdminRotation ran, and the new admin DID came back in the reply. No orphan ACL entry was left. So the VTA and the grant are fine; the failure is in TSP reply delivery.
Same fault in the VTC setup wizard
vtc setup(vtc-service 0.11.58, in OpenVTC/verifiable-trust-infrastructure) failed the same way on 2026-10-10, with a 164 KB reply. The same one-line workaround applies todrive_provisioninvtc-service/src/setup/wizard.rs. The runner is vta-sdk'sprovision_client, so suggestions 1 and 2 may belong there rather than in each wizard. Related earlier work there: OpenVTC/verifiable-trust-infrastructure#766, OpenVTC/verifiable-trust-infrastructure#767, OpenVTC/verifiable-trust-infrastructure#869.All reactions