Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Claims review

Every public text about FerroEHR is read, once per release, against EHDS Art. 28 (Regulation (EU) 2025/327). The article reads: “In the information sheet, instructions for use or other information accompanying EHR systems, and in the advertising of EHR systems, it shall be prohibited to use text, names, trademarks, pictures and figurative or other signs that may mislead the professional user […] with regard to their intended purpose, interoperability and security by:

  • (a) ascribing functions and properties to the EHR system which it does not have;
  • (b) failing to inform the professional user of likely limitations related to interoperability or security features of the EHR system in relation to its intended purpose;
  • (c) suggesting uses for the EHR system other than those stated to form part of the intended purpose in the technical documentation.“

This page is the record of those reviews: what was read, what was found, what was changed. It also carries the re-assessment of EHDS Annex II 2.5. The measure for point (c) is the intended-purpose statement.

Warning

The quotations are from the Official Journal text vendored at docs/law/eu/ehds/text.html (EUR-Lex). Art. 28 takes “professional user” from Article 3, point (8), of Regulation (EU) 2018/1807, which the repository does not carry, so these pages do not restate that definition. A text that passes this review has been read against Art. 28 by the manufacturer; no authority has examined it.

What is reviewed, and how

The texts in scope, each release:

  • the repository README.md and the landing page at ferroehr.eu;
  • every page of this book, except the pages generated from the tracker and from committed measurement records, whose generators are reviewed instead;
  • SECURITY.md;
  • the READMEs of the published openehr-* crates;
  • the OCI labels of the three images (docker/Dockerfile, docker/viewer/Dockerfile, docker/postgres/Dockerfile) and the Helm chart’s description;
  • the public sandbox at sandbox.ferroehr.eu.

Each claim about a function, a property, interoperability or security is checked against the code, the configuration defaults and the committed records, and against the intended-purpose statement. A misleading claim is fixed in the same change as the review. Measured numbers are never typed into a text: the conformance and performance figures come from generators over the committed run records, and a CI check fails a page whose numbers disagree with them.

The review of 2026-10-06

Read on the tree after commit d6539d470, before FerroEHR 4.3.5.

Changed

#TextClaimArt. 28FindingNow reads
1README, opening“Every compliance claim it makes is machine-verified”(a)Only the openEHR conformance numbers are machine-verified. The regulatory pages are reviewed by hand, and “compliance” invites the regulatory reading“Every conformance claim it makes is machine-verified”
2README, “What makes it different”“Compliance you can verify.”(a)as row 1“Conformance you can verify.”
3README, licence paragraph and “Security & operations”“Multi-tenancy […] fully integrated: each tenant is an isolated logical openEHR system, enforced by PostgreSQL row-level security”(a)FerroEHR is single-tenant, with no tenant column, row policy or tenancy configuration (one instance, one organisation)“One instance per organisation: FerroEHR is single-tenant […]”; multi-tenancy removed from the feature list
4README, “Security & operations”national identifiers “sealed under a per-tenant key”(a), (b)there are no tenants, and the sealing is off by default“sealed under a key the deployment holds […] (off by default)”
5README, “Security & operations”“IHE ATNA-compliant system log (DICOM audit messages over TLS syslog)”(a), (b)“compliant” claims a property nobody assessed; the syslog sink is off by default and ships with UDP; the default sink is the local hash-chained storethe access log, its two renderings, the local store and the optional forwarding, described as they ship
6README, “Security & operations”“Hardened by default”(b)TLS is off by default, and the shipped deployment.profile is sandbox, which must not hold real patient data“Hardening built in”, with both limitations stated
7README, “Integration”the FHIR R4B connectors, with no limitation(b)the in-tree connector was removed in favour of FerroBRIDGE in v4.3.6 (#3080)the removal is stated
8Landing page, title, og: and twitter: descriptions, structured data“Spec-compliant and measured”, “spec-conformant and measured”(a)states conformity as a property; what exists is a measured openEHR conformance record“openEHR conformance, measured”
9Landing page, hero note“every compliance claim machine-verified”(a)as row 1“every conformance number machine-verified”
10Landing page, conformance section“Compliance you can verify.” and “Every release runs the complete conformance catalogue against the live server”(a)the release pipeline does not run the catalogue; the committed record measured FerroEHR 4.3.0“Conformance you can verify.”; the record “names the FerroEHR version it measured”
11Landing page, feature grid heading“Everything the openEHR standard asks for”(a)claims complete coverage of the standard, which nothing measures“The openEHR platform services”
12Landing page, Security card“multi-tenancy”(a)as row 3“One instance per organisation.”
13Book, Introduction“Compliance you can run yourself”(a)as row 1“Conformance you can run yourself”
14Book, System architecture“why the compliance claims are checkable”, “the compliance claims are machine-derived”(a)as row 1“conformance claims”
15Book, Beyond the core“Security, multi-tenancy, and the audit trail”(a)as row 3“Security, instance separation and the audit trail”

Read and kept

  • The conformance and performance figures on the README, the landing page and the book: rendered from docs/conformance/ferroehr/ by scripts/render/conformance-stats.sh and scripts/render/perf-assets.sh, and checked in CI. No number is typed.
  • “openEHR-spec-conformant”, on the Conformance page, in the openehr-* crate READMEs, in the org.opencontainers.image.description label of the server image and in the Helm chart description: a claim about the openEHR specifications, backed by the committed record (CORE and STANDARD, every case passed) and explained on that page. It makes no claim about any regulation.
  • “aims to be the first openly developed, source-available openEHR CDR with a published, tracker-backed EU compliance posture” (the compliance overview, Why FerroEHR exists): stated as an aim, beside the statement that FerroEHR holds no certification and no declaration.
  • The compliance pages, the DPIA and the go-live checklist: each says it does not make a deployment compliant with anything.
  • The image labels: the server image carries eu.ferroehr.image.carries-sandbox-posture="true", which tells an operator the shipped profile; the manufacturer label names Cadasto B.V.
  • SECURITY.md: its commitments match the procedures in docs/post-market.md and the release procedure.
  • The public sandbox: the viewer shows the sandbox notice on every page, the README calls its content demo data, wiped nightly, and the landing page calls it a live demo.

Left open

  • The Helm chart’s description (Chart.yaml, and the chart README that helm-docs generates from it) says “hardened-by-default security posture” and lists the workload hardening, which is accurate, but does not say that a stock install runs the sandbox profile. The chart now declares config.deployment.profile: sandbox in its values, and the chart README says so under “Before you install”; the Chart.yaml description still does not.

The review of 2026-10-08

Read on the tree after commit de648fcc3, for FerroEHR 4.3.6. This release removes the in-tree FHIR connector (#3080) and moves the deployment gaps off the public status document (#3671), so those two facts were checked in every text in scope.

Changed on 2026-10-08

#TextClaimArt. 28FindingNow reads
1Landing page, “Fits your stack”“FHIR R4B”(a)FerroEHR maps no FHIR resources; it reaches FHIR R4B terminology servers and renders its access log as FHIR R4 AuditEvent“FHIR R4B terminology”
2Book, instructions for use, CRA user information, configuration, go-live checklistan accepted deployment gap is stated “on GET /ferroehr/rest/status”(a)the gaps moved to the authenticated GET /management/status (#3671); the public status document carries the profile only“on GET /management/status”
3Book, instructions for use, the deployment profilethe list of what production refuses(b)omitted the plaintext audit feed and the plaintext object store (#3676, #3689)both named
4Book, instructions for use, limitationsnone on FHIR(b)an operator of an earlier release may rely on the removed connectorFerroEHR maps no FHIR resources, and [fhir] is refused at boot
5Book, information sheet, (e)the standards table(b)omitted the FHIR terminology operations and RFC 9470, which this release usesboth listed

Read and kept on 2026-10-08

  • The README’s integration list and the book’s Beyond the core: both state that FHIR R4 mapping is FerroBRIDGE’s and name the two uses of FHIR FerroEHR keeps.
  • Single tenancy: the README, the landing page and the book state one instance per organisation; “multi-tenant” appears only where the book quotes openEHR’s placement of multi-tenancy outside one system.
  • The Helm chart’s description: it now says a stock install runs the sandbox profile (#3637), which closes the item left open on 2026-10-06.

Left open on 2026-10-08

Nothing.

The review of 2026-10-15

Read on the tree after commit 606a29cce, for FerroEHR 4.3.9. The facts of this release checked in every text in scope: the ABAC patient scope over EHR-only queries and [authz.abac] patient_namespace (#3771); a restricted EHR_ACCESS default with the role gate off, and the rbac_off gap (#3742); national identifiers sealed per party version, the sealed form in write responses and the refusal of plaintext protected values on import (#3772); digest version 2 of the audit chain (#3686); the AQL row rules (#3733, #3734); the removal of FOSSA; the TLS certificate reload and EC keys (#3765, #3761); the four-cluster production layout (#3755 to #3757, #3762); the example generator (#3739, #3740); quoted GET query parameters (#3781). The 2026-10-08 move of the deployment gaps to GET /management/status (#3671) was re-checked on every page, and eleven pages still named the public document.

Changed on 2026-10-15

#TextClaimArt. 28FindingNow reads
1Book, licensing, security, go-live checklist (two items), Kubernetes, authentication configuration (two places), Compose (two places), installation, configuration, command line, DPIA R9, compliance overview (GDPR Art. 25, DSV Art. 4)an open or accepted deployment gap, or the licence in force, is reported on GET /ferroehr/rest/status (or /rest/status)(a)the public document carries status, the versions, the timestamp and the profile only; the gaps, the licence and the support period are on GET /management/status (#3671)“on GET /management/status”; the configuration page adds that the public document carries the profile only
2Book, instructions for use, the deployment profilethe list of what production refuses(b)omitted the role gate off while authentication is on, the rbac_off gap this release adds (#3742)named
3Book, instructions for use, limitationsnone on how AQL reads repeated nodes(b)AQL 1.1.0 does not define the rows of a path through a repeated node; this release changed them (#3733), so another server, or a client written for an earlier release, may count differentlythe limitation is stated, with the regrouping a client needs
4Book, instructions for use, limitationsnone on the audit chain’s coverage(b)records written before the upgrade keep digest version 1, which leaves out the columns added after the chain was introduced (#3686)stated, with the columns and the trigger that alone protects them
5Book, configuration, migrate_on_runtime_credentiala domain relocated to its own database “is prepared on that domain’s runtime DSN”(a)[database.<domain>] migrate_url` now prepares it on its own migration credential (#3756); only without it does the runtime DSN prepare it and open the gapboth cases stated
6Book, records of processing, categories of personal dataclinical content “in the ehr schema”; parties “in the demographic schema”(a)the schemas are clinical and party since the storage rewriteclinical and party
7Book, Kubernetes, server TLSthe viewer’s OIDC client “trusts the public roots it is built with”, so a private-CA issuer is not trusted(a)the viewer’s OIDC client verifies against the platform trust store (#3767)a private-CA issuer is trusted once its CA is in the store or under SSL_CERT_DIR
8Book, querying with AQL, the URL formhow a parameter value binds(b)silent on a quoted value, which now binds the JSON string it denotes (#3781)the quoted form and the malformed case stated
9Book, templates and validation, the example composition“a skeleton instance you can fill in”, with no limitation(b)the example passes the server’s own validation except for two open defects (#3779, #3780) and templates whose mandatory content only a slot admitsthe exceptions named
10Book, threat model, what each database credential can reachthe writer roles hold SELECT, INSERT, UPDATE and DELETE on their schema(b)the migration bookkeeping is now read-only to every runtime role, and the boot refuses a writable onestated below the table
11Book, information sheet, (e)TERM “3.1.0”(b)the server applies amendment 3.1.4 (SPECPR-450), which adds properties and units the template upload and the quantity validation accept (#3717, #3727)“3.1.0, with amendment 3.1.4 (SPECPR-450)”

The technical documentation’s Annex III 2 named the public status document for the deployment posture too; it names GET /management/status now. | 12 | Repository README, the template-model and example bullets (two places) | every template “generates spec-valid example compositions that pass the server’s own validation” | (a) | #3779 and #3780 are open, and a template whose mandatory content only an archetype slot admits has no committable example | “a template […] generates” them, “except for the cases the book’s example page lists” | | 13 | Repository README, the deployment posture and licence bullets | sandbox names every open gap “on GET /rest/status”; “GET /ferroehr/rest/status reports the grant in force” | (a) | both are on GET /management/status (#3671) | “on GET /management/status” | | 14 | Book, EHDS readiness row 1.2 | its issue reference | (a) | the probe that issue asked for ran (#3692), without hostUsers: false or sslmode=verify-full | names #3788 for those two |

Read and kept on 2026-10-15

  • FOSSA: no text in scope names FOSSA, a FOSSA badge or a FOSSA lane after its removal in this release.
  • The patient scope and patient_namespace: the [authz.abac] table and its patient-scope list state the EHR-only queries, the direct-route refusals and the namespace rule as they ship; the security page’s fail-closed box is unchanged by them.
  • The restricted EHR_ACCESS default and rbac_off: the authentication configuration, the configuration page, the go-live checklist, the Kubernetes values table and the CRA risk assessment (d) state both.
  • Identifier protection: the privacy configuration, the threat model’s sealed-identifier section, the operations guide and the admin API page state the per-version rows, the sealed write response, the plaintext refusal on import and load, and the erasure of sealed rows.
  • The audit chain: the audit trail page states both digest versions and the downgrade finding.
  • AQL: querying with AQL and the AQL engine concept page state the row rules and say AQL 1.1.0 does not define them.
  • TLS: the server configuration and the Kubernetes page state the reload interval, the EC key formats and the explicit-curve refusal.
  • The information sheet’s BALP row (“the profile revision is not pinned”): kept. The delete agent types follow BALP 1.1.4, and the profile URLs the records carry are unversioned.
  • The image labels and the Helm chart’s description: unchanged by this release but for the chart description’s NetworkPolicy defaults (#3743), which match values.yaml.

Left open on 2026-10-15

Nothing.

The review of 2026-10-16

Read on the tree after commit e923076de, for FerroEHR 4.3.10. The facts of this release checked in every text in scope: the usage report in format 2 (#3795); the usage report off in a CI run, whatever the switch says (#3794); the dump and load archive on zip 9.0.4 (#3798).

Changed on 2026-10-16

#TextClaimArt. 28FindingNow reads
1Book, instructions for use, the usage report“Every deployment sends a daily usage report to Cadasto B.V. by default”(a)the report also goes out at each start, and a server started in a CI run sends none (#3794)both stated, with the two CI variables
2Book, operations, the usage report“It is on by default”(a)off in a CI run, whatever the switch says (#3794)“on by default outside a CI run”

Read and kept on 2026-10-16

  • The usage report’s fields: the usage report page lists the format 2 fields and the status object as the server sends them, and says earlier releases send format 1. The field list in the CRA risk assessment’s outbound data flows, the shared responsibility summary, the records of processing note, the intended purpose and the threat model name the same information, which format 2 only regroups.
  • The CI switch: the usage report page states the two variables, the boot line and the --print note as they ship, and the server configuration table states the two variables.
  • The dump and load archive: the admin APIs page names no zip library and its compression_format values are unchanged.
  • The README, SECURITY.md, the crate READMEs, the image labels and the Helm chart’s description: unchanged by this release, and none names the usage report’s format or the zip library.

Left open on 2026-10-16

Nothing.

The review of 2026-10-19

Read on the tree after commit a15290689, for FerroEHR 4.3.11. The facts of this release checked in every text in scope: the viewer drawn in FerroUI 0.1.6 (#3802, #3811, #3815, #3820); the viewer’s tab icon (#3808); the coming split of the logging component with FerroSYS (#3662); the zip advisories judged not affecting in the Rust-dependency VEX document (#3806).

Changed on 2026-10-19

Nothing. Every text in scope already states these facts.

Read and kept on 2026-10-19

  • The viewer pages: the viewer names the footer’s manufacturer line and the System screen’s Manufacturer card (#3634), and dark mode describes the moon and sun button and the theme cookie as FerroUI draws them. No page names a component library the viewer no longer uses.
  • The logging component: the EHDS readiness page and the technical documentation say FerroEHR ships the whole logging component today and that FerroSYS takes the central store, review, retention and retrieval in a later release; nothing calls that split shipped.
  • The security notices: the release’s Security entry names the VEX document that records the three zip advisories as not affecting FerroEHR.

Left open on 2026-10-19

Nothing.

The review of 2026-10-19, for 5.0.0

Read on the tree after commit ec78e7f40, for FerroEHR 5.0.0. The facts of this release checked in every text in scope: openehr-sdt builds a canonical PARTY from a party Web Template and stamps the caller’s rm_version (#3810, #3809); the openEHR terminology serves the interoperability group; archetype identifiers follow the BASE 1.3.0 grammar under the development profile (#3835, #3845); the nine openehr-* crates at 0.0.89; only the newest release is supported (#3853); the viewer serves its hydration bundle on a route of its own (#3849); the viewer draws with FerroUI 0.1.7 and chooses its language from Accept-Language or [ui] locale.

Changed on 2026-10-19, for 5.0.0

  • The Rust crates: the dependency examples named the crates at 0.0.86, three versions behind the published line; they name 0.0.89. The openehr-sdt row names the PARTY build beside the composition formats.
  • The viewer: the ui.locale row said a deployment adds a language as a catalog file. The catalogs are embedded when the viewer is built, so the row now says a new catalog reaches a deployment in a rebuilt image, never at run time.

Read and kept on 2026-10-19, for 5.0.0

  • The support period: SECURITY.md, the Rust crates page and the release notes say that only the newest release is supported and that the five years of security fixes are a commitment for the product, shipped in the newest release. No text in scope still promises fixes to an earlier release.
  • The crate READMEs: the openehr-sdt README describes the composition formats and does not mention the PARTY build. An omission ascribes no function the crate lacks, and the README changes with the crate’s next content version.
  • The specification pins: docs/VERSIONS.md and the crate table name BASE 1.2.0 and 1.3.0, which is what the two generations implement.

Left open on 2026-10-19, for 5.0.0

Nothing.

The review of 2026-10-20, for 5.1.0

Read on the tree after commit 18097c6f1, for FerroEHR 5.1.0. The facts of this release checked in every text in scope: authentication and the SMART on openEHR scope check run on the shared ferrohealth-smart crate, only asymmetric algorithms verify a token, the HMAC key is gone, and every route requires a scope (#3872); the Keycloak overlay and the development stack serve their issuer over TLS with a development CA, and the CDR accepts their tokens again (#3873); AQL rooted at PERSON, ORGANISATION and ROLE, and a demographic root joined to its EHR’s content (#3818, #3831); the audit FHIR feed and the remote policy service authenticate with OAuth 2.0 client credentials (#3829); the chart mounts a database CA for sslmode=verify-full (#3788); a rotated mTLS client CA is re-read without a restart (#3769); the viewer reads every word from its catalog with FerroUI 0.2.0 (#3860); the nine openehr-* crates at 0.0.90.

Changed on 2026-10-20, for 5.1.0

  • The repository README.md: the authentication bullet named Basic and OAuth2/OIDC alone. Every route now refuses a caller whose scopes do not grant the operation, a limitation a deployment must know (Art. 28(b)), so the bullet says the SMART on openEHR scopes are checked on every route.
  • The Rust crates: the dependency examples named the crates at 0.0.89; they name 0.0.90.

Read and kept on 2026-10-20, for 5.1.0

  • The Keycloak quickstart: Docker Compose, getting started and the README offer the overlay for bearer-token authentication. Between #3872 and #3873 the CDR did not accept the overlay’s tokens; on this tree it does, and the compose page names the development CA and the scopes the realm puts in every token.
  • Authentication and authorization: security, the authentication reference and SMART App Launch describe the [auth] keys, the RFC 9068 profile and the scope check as the code does.
  • Annex II 2.5: a caller refused for want of a scope is refused by the grant its own identity provider issued, which the deployment controls; the re-assessment below stands.

Left open on 2026-10-20, for 5.1.0

Nothing.

Annex II 2.5, re-assessed

EHDS Annex II 2.5: “The harmonised software components of an EHR system shall not include features that prohibit, restrict or place an undue burden on authorised access, personal electronic health data sharing or use of personal electronic health data for permitted purposes.” The text does not define “undue”.

The EHDS readiness page marks the row “Shipped” with two pieces of evidence: the query surface over the whole stored record, and an authorisation layer that refuses or permits without a commercial gate. The question re-assessed here is whether the authorisation defaults change that, now that production refuses the open EHR_ACCESS default unless it is accepted by name.

  • What “authorised” means here. The deploying organisation decides who is authorised, and expresses it in RBAC roles, ABAC policy and per-EHR EHR_ACCESS settings. A refusal of a caller the organisation has not authorised is not a burden on authorised access.
  • The restricted default. Under ehr_access_default = "restricted", an EHR with no settings is reachable only by the admin role, so a professional the organisation means to authorise is refused until someone writes that EHR’s settings. That is a burden, and it is the organisation’s choice: the open default remains available, production accepts it by name, and the restrictive one exists because GDPR Art. 25(2) asks that personal data are not by default “made accessible without the individual’s intervention to an indefinite number of natural persons”. The settings take role entries (role:nurse), so one entry covers a whole role, and the admin role can always reach the record to write them (per-EHR access control).
  • Restriction of processing. A record under restriction is refused on every read path. The controller sets the mark under GDPR Art. 18; it is a restriction the organisation decides on, not one the software imposes (retention, restriction and objection).
  • Small-cell suppression. A cohort query that would serve fewer distinct EHRs than privacy.cohort.small_cell_threshold (default 5) withholds its rows and says so. It applies only to population queries across the pseudonymisation boundary, and the deployment sets the threshold.
  • Rate limits and the admission limit. On by default, set above the server’s own measured whole-server ceiling, so they refuse abuse and leave normal load alone (request limits).
  • The licence. The licence in force changes only the stamp on new identifiers; it gates no feature, endpoint, limit or message (app/ferroehr/src/licence/mod.rs).

Outcome. No feature found prohibits or restricts access, sharing or use that the deploying organisation has authorised, and the burdens found are settings the organisation chooses and can change. The “Shipped” status stands on this reasoning, which the readiness page will cite.

Review record

One row per release. The release procedure adds the row when the release is cut, after the texts in scope have been read.

ReviewedTreeChangedLeft open
2026-10-06before 4.3.5, after d6539d47015 claims, in the README, the landing page and three book pagesthe chart description (#3637)
2026-10-084.3.6, after de648fcc35 claims, in the landing page and five book pages; the chart description item closednothing
2026-10-154.3.9, after 606a29cce14 claims, in 17 book pages, the README and the readiness rownothing
2026-10-164.3.10, after e923076de2 claims, in two book pagesnothing
2026-10-194.3.11, after a15290689nonenothing
2026-10-195.0.0, after ec78e7f402 claims, in two book pagesnothing
2026-10-205.1.0, after 18097c6f12 claims, in the README and one book pagenothing
2026-10-205.2.0, after e5aa1902b1 claim, in one book pagenothing
2026-10-225.3.2, after 434cd351e4 claims, in four book pages and the technical documentationnothing
2026-10-235.3.3, after 25845e6d31 claim, in one book page: the templates page said `raw` is write-only, corrected for #3934
2026-10-245.4.0, after de38ec433nonenothing
2026-10-245.4.1, after 9fdd360912 claims, in two book pages: the compliance index named [terminology.external] and a terminology route, and the crates page pinned 0.0.97nothing
2026-10-255.4.2, after 2333f98banone; the changes of this release updated their own pages in the pull requests that made themnothing
2026-10-275.5.0, after bed3543e31 claim, in the hazard log: H2 named a middleware test renamed by #4192the conformance chapter bars counted the CNF TDD cases until the Veredictum 0.3.3 record, which carries none, replaced them (#4204)
2026-11-045.6.0, after 25db5c69c4 claims, in the README, the landing page and two book pages: each said FerroBRIDGE’s broker delivery was in progress, which FerroBRIDGE 0.0.15 shipsthe canonical-XML reverse_relationships left open by the 5.6.0-rc8 review shipped (#4778); a v2-namespace OPT 1.4 ordinal list stays tracked on #5258
2026-11-055.6.1, after 75262e1ce1 claim, in the Kubernetes installation: it said the chart pins FerroPIX 0.0.15, which the 5.6.1 pins moved to 0.0.16a v2-namespace OPT 1.4 ordinal list stays tracked on #5258
2026-11-055.6.2, after ea1190926none; the texts in scope make no claim this release changesa v2-namespace OPT 1.4 ordinal list stays tracked on #5258
2026-11-055.6.3, after 6b2ff1644none; the texts in scope make no claim this release changesa v2-namespace OPT 1.4 ordinal list stays tracked on #5258

The review of 2026-10-20, for 5.2.0

Read on the tree after commit e5aa1902b, for FerroEHR 5.2.0. The facts of this release checked in every text in scope: authentication runs on ferrohealth-smart 0.3.1, with [auth.basic] cache_ttl_seconds in place of [authz] basic_cache_ttl_seconds, scopeless_subjects accepted and a malformed bearer token answered 401 (#3875); [auth.smart] issuer_metadata_file (#3875); the EHDS Art. 8 restriction of access with its emergency override (#3682); the audited read of a party’s protected identifiers (#3776); an EHR extract never carries a sealed identifier (#3777); the FLAT history origin from the earliest event (#3888); the nine openehr-* crates at 0.0.91.

Changed on 2026-10-20, for 5.2.0

  • The Rust crates: the dependency examples named the crates at 0.0.90; they name 0.0.91.

Read and kept on 2026-10-20, for 5.2.0

Left open on 2026-10-20, for 5.2.0

Nothing.

The review of 2026-10-22, for 5.3.2

Read on the tree after commit 434cd351e, for FerroEHR 5.3.2. The facts of this release checked in every text in scope: the access log written through FerroAUDIT’s ehds-logging and ihe-atna 0.2.4, with the categories named by their HL7 Europe codes, category_basis, and the record’s jti, operation subtype and outcomeDesc written again (#3887, #3902); the retention register keyed on the category token (clinical migration 0016); the syslog feed over RFC 5425 TLS only, with the UDP transport removed; ferrohealth-smart 0.4.0 and one login record per Basic credential verification (#3895); a + in a query string read as a space (#3891); the subject id of a lookup by subject kept out of every log, span and error body (#3906); Rust 1.99.1 with the MSRV at 1.98; the nine openehr-* crates at 0.0.95.

Changed on 2026-10-22, for 5.3.2

  • The Rust crates: the dependency examples named the crates at 0.0.91; they name 0.0.95.
  • EHDS readiness, the European logging component row (from website/book/ehds.yaml): it named ehds-logging and ihe-atna at 0.1.1; it names 0.2.4, the version the root Cargo.toml pins.
  • The hazard log, the component paragraph: as the row above.
  • The technical documentation’s scope (docs/technical-documentation/README.md): as the row above.

Read and kept on 2026-10-22, for 5.3.2

Left open on 2026-10-22, for 5.3.2

Nothing.

The review of 2026-10-24, for 5.4.0

Read on the tree after commit de38ec433, for FerroEHR 5.4.0. The facts of this release checked in every text in scope: the viewer rebuilt on FerroUI 0.3.6 in islands mode, working with JavaScript off, with its new addresses, its preferences as cookies of their own and an open tab that idles out (#3932, #3057); an ECDSA or Ed448 OpenPGP signing key refused at boot for RUSTSEC-2026-0387 (#3965); the signer’s choice of subkey and primary key (#3967, #3969, #3971); the audit drain closing its syslog connection with close_notify and FerroAUDIT 0.2.5 (#3959); the nine openehr-* crates unchanged at 0.0.97.

Changed on 2026-10-24, for 5.4.0

Nothing: the pull requests that made these changes updated the texts they touched.

Read and kept on 2026-10-24, for 5.4.0

  • The viewer statements: the README, the viewer pages and the image label of docker/viewer/Dockerfile describe the viewer the code draws; no text names a removed address, the ferroehr-theme cookie, the Tailwind build or a background poll.
  • The signing statements: PGP signing and the authentication and signing configuration state the refused ECDSA and Ed448 keys, the subkey the server signs with and the retired-key path, as the code does.
  • The audit statements: the hazard log and EHDS readiness name ehds-logging and ihe-atna 0.2.5, the version the root Cargo.toml pins.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the other image labels and the chart description make no claim this release changes.

Left open on 2026-10-24, for 5.4.0

Nothing.

The review of 2026-10-24, for 5.4.1

Read on the tree after commit 9fdd36091, for FerroEHR 5.4.1. The facts of this release checked in every text in scope: the CDR asks one terminology server under [terminology] through ferrobridge-term, and chart 14.0.0 wires it (#3085); an unserved method is 405 with Allow under SMART (#3987); the conformance statement declares unmatched-member-refused (#3988); a ROLE template whose single attributes state 0..* uploads, and a demographic template’s example is a party (#3994, #3995); the viewer offers demographic templates only where they work (#3997); the nine openehr-* crates at 0.0.98.

Changed on 2026-10-24, for 5.4.1

  • The compliance index (SGB V § 355 row) named [terminology.external] and “the terminology route”; it now names [terminology] and its url.
  • The crates page pinned the openehr-* crates at 0.0.97; this release publishes 0.0.98.

Read and kept on 2026-10-24, for 5.4.1

  • The terminology statements: the integrations configuration, the Kubernetes installation and the chart README describe the one [terminology] and secrets.terminologyClientSecret, as the code and the chart do.
  • The template statements: the templates page and content negotiation state the example party of a demographic template and its JSON and XML forms.
  • The 405 statement: content negotiation’s “405 always names the allowed methods” holds under SMART as well.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-24, for 5.4.1

Nothing.

The review of 2026-10-25, for 5.4.2

Read on the tree after commit 2333f98ba, for FerroEHR 5.4.2. The facts of this release checked in every text in scope: a valid Certification Revocation of a Direct Key signature cancels it, so it no longer lets the primary key sign (#3972); an access record whose patient lookup failed writes unavailable (#3900); the viewer’s Paths tab offers a query only on a COMPOSITION template, on FerroUI 0.3.7 (#4001); ferrohealth-smart 0.5.0, one openehr-sdt in the build (#4002); ihe-atna and ehds-logging 0.2.7 (#3954); zip 8.6.0 out of the build (#3799); chart 14.0.1; the nine openehr-* crates at 0.0.98.

Changed on 2026-10-25, for 5.4.2

Nothing at the cut: each change above updated its page in its own pull request.

Read and kept on 2026-10-25, for 5.4.2

  • The signing statement: the PGP signing page states that a revoked Direct Key signature is not chosen, as the code does.
  • The viewer statement: Templates & EHRs says a node offers Build a query on this path only on a COMPOSITION template.
  • The logging component: the EHDS readiness page, rendered from the FerroHEALTH book’s website/book/ehds.yaml, and the hazard log name ehds-logging and ihe-atna 0.2.7.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 14.0.1 and image 5.4.2.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-25, for 5.4.2

Nothing.

The review of 2026-10-27, for 5.5.0

Read on the tree after commit bed3543e3, for FerroEHR 5.5.0. The facts of this release checked in every text in scope: the server and the viewer run on the family’s ferrohealth-server, ferrohealth-config, ferrohealth-telemetry, ferrohealth-smart, ferrohealth-authz and ferrohealth-secret crates, with the renamed and retired configuration keys under Changed and Removed (#4087, #4038, #4028, #4139, #4214); the EHR index asks an IHE patient identity source over PIXm by default (#3380); opaque access tokens are checked by introspection (#4033); [auth] ca_file trusts a private issuer CA (#3909); the audit trail is delivered and retrieved through ihe-atna and carries the data permit (#4029, #4030, #4120); Template Data Document import is removed (#4199); the quickstart Keycloak overlay trusts its CA through ca_file and signs the viewer in (#4236, #4243); the chart’s usageReport.enabled=false reaches FerroTERM, FerroBRIDGE and FerroPIX (#4270); chart 20.0.0 with appVersion 5.5.0.

Changed on 2026-10-27, for 5.5.0

  • The hazard log: H2 named ehr_id_from_various_paths, which #4192 replaced with ehr_id_from_the_matched_route; the log names the new test.

Read and kept on 2026-10-27, for 5.5.0

  • The configuration pages: the server configuration, authentication, security and the viewer name the new keys and say which old keys refuse the start (slow_aql_ms, user_role, LEPTOS_SITE_ADDR); no page still offers token_url, server.bind, ferroehr config default, healthcheck --url or the TDD routes.
  • The CRA risk assessment: its outbound data flows table gains the token introspection flow and the patient identity source over PIXm, with revision 14.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 20.0.0 and image 5.5.0, and the compose page names FERROEHR_CONFIG and the Keycloak overlay’s ca_file.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-27, for 5.5.0

  • The conformance chapter bars showed the CNF TDD cases from the last measured run until the Veredictum 0.3.3 record, which carries none, replaced the record and the bars (#4204).

The review of 2026-10-28, for 5.6.0-rc1

Read on the tree after commit df14b4311, for the FerroEHR release candidate 5.6.0-rc1. The facts of this release checked in every text in scope: the REST surface serves three namespaces, ITS-REST alone under /openehr/v1/, FHIR and IHE under /fhir/r4/ and FerroEHR’s extensions under /extensions/v1/, with no redirect from the old paths (#4330); the audit store renders each record’s AuditEvent from its columns, answers every ITI-81 parameter through search tokens, and its migration refuses a store holding records of an earlier release (#4300, #4302, #4313); ferrohealth-server 0.3.7 bounds a silent connection by [server] header_read_timeout_ms (#4283, GHSA-2mc2-h3jq-c5wg); an AQL query nested past its depth bound answers 400 (#4284); a cut or unreachable database answers 503 with Retry-After within about a second (#4290, #4323); an abandoned query frees its connection (#4285); storage generation 3 (#4306); a deployment that declares no category map records accesses through the map FerroEHR ships (#4298); chart 20.0.1-rc1 with appVersion 5.6.0-rc1.

Changed on 2026-10-28, for 5.6.0-rc1

  • The hazard log: H7 named the_current_digest_covers_every_column and records_under_the_original_digest_still_verify and kept a remainder for records under the version 1 digest; #4300 retired the versioned digests with the records they covered, so H7 names the_digest_covers_every_column and the render-policy and search-token tests, and drops that remainder. H3 said FerroEHR ships no category map; it names the shipped map and its test (#4298).
  • The technical documentation named the FerroAUDIT pins as 0.2.5; it names 0.3.1, the version the root Cargo.toml pins.

Read and kept on 2026-10-28, for 5.6.0-rc1

  • The API pages: the API overview names the three namespaces, the resources page the paged CONTRIBUTION list under /extensions/v1/, and the audit trail the ITI-81 route under /fhir/r4/ and the empty audit store 5.6.0 starts on; no page in scope names a moved route under /openehr/v1/.
  • The operations pages: AQL states the depth bounds, operations the 503 on a cut or unreachable database, and the server configuration and security header_read_timeout_ms.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 20.0.1-rc1 and image 5.6.0-rc1 in the book of main; the frozen book of 5.5.0 under /docs/latest/ keeps the 5.5.0 pins, since a pre-release freezes no book.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-28, for 5.6.0-rc1

  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths until 5.6.0 ships.

The review of 2026-10-29, for 5.6.0-rc2

Read on the tree after commit 40982f374, for the FerroEHR release candidate 5.6.0-rc2; 5.6.0-rc1 was tagged at 61d5d1ff7 and not published. The review of 2026-10-28 stands for everything 5.6.0-rc1 carried. The facts added since and checked in every text in scope: the [[authz.origin]] register stamps a direct commit from a registered client with that client’s origin, a retention period may be keyed on a registered origin, and the ABAC request carries category and origin attributes (#4299); a client CA under client_auth = "none" is refused at load (#4349, GHSA-52mf-qg39-4q4j); the chart refuses to render without an explicit deployment profile (#4294); the logging component’s intended performance is stated and measured (#4293) and its ATNA transactions are exercised against independent implementations (#4296); an unclassified record’s FHIR rendering carries the BALP entity:data slice (#4353); chart 21.0.0-rc2 with appVersion 5.6.0-rc2.

Changed on 2026-10-29, for 5.6.0-rc2

  • The hazard log: H4 said a body without FEEDER_AUDIT records this server as its origin; it names the registered client’s origin and its test (#4299). H3’s remainder said the chart’s empty map matches the binary, which ships a map since #4298; it says the shipped map applies.

Read and kept on 2026-10-29, for 5.6.0-rc2

  • The retention and access pages: retention, the audit configuration and security describe the [[authz.origin]] register, the origin-keyed retention policy and the two ABAC attributes as the code implements them.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc2 and image 5.6.0-rc2 in the book of main, and the Kubernetes page sets config.deployment.profile in its install example (#4294).
  • The TLS and performance pages: the server configuration says client_ca_file is refused under client_auth = "none", and logging performance states the claims scripts/perf/audit.sh measures.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-29, for 5.6.0-rc2

  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths until 5.6.0 ships.

The review of 2026-10-29, for 5.6.0-rc3

Read on the tree after commit df26a92e6, for the FerroEHR release candidate 5.6.0-rc3. The review of 2026-10-29 for 5.6.0-rc2 stands for everything that release candidate carried. The facts added since and checked in every text in scope: every error on the openEHR, extension and management routes is RFC 9457 problem details, the violations of a validation refusal reach a client as errors under Prefer: return=representation, and an authentication refusal under the FHIR base is an OperationOutcome (#4342); the openehr-its client reads a problem details refusal (#4355); a + in a query string is a literal plus (#4356); FROM PERSON p CONTAINS EHR e joins through the EHR’s EHR_STATUS.subject.external_ref (#4411); a projection ending on a list attribute answers the list once (#4413); the local EHR index is a supported posture (#4379); [server] idle_timeout_ms closes an idle kept-alive connection on ferrohealth-server 0.4.0 (#4417); the FHIR AuditEvent names the object it touched by entity.what.identifier on FerroAUDIT 0.3.2 (#4366); openehr-query is Apache-2.0 from 0.0.116 (#4405); the chart and the sandbox run FerroPIX 0.0.12 (#4392); chart 21.0.0-rc3 with appVersion 5.6.0-rc3.

Changed on 2026-10-29, for 5.6.0-rc3

  • The validation pages: template validation and the privacy configuration said a 422 lists its violations in validationErrors, the body #4342 replaced; they name the problem’s errors, sent under Prefer: return=representation, and link the validation-failed type.
  • The hazard log: H2 states that the FHIR AuditEvent names the object it touched by entity.what.identifier and names its test (#4366).

Read and kept on 2026-10-29, for 5.6.0-rc3

  • The API pages: Problem types and content negotiation and errors describe the problem body, its types and the Prefer members as the code writes them; querying with AQL describes the party join through the subject reference and the list projection as the engine answers them.
  • The server configuration: the server configuration states idle_timeout_ms and the head bound as ferrohealth-server 0.4.0 applies them.
  • The licensing page and the crate READMEs name openehr-query Apache-2.0 from 0.0.116 and BUSL-1.1 for 0.0.60 to 0.0.115.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc3 and image 5.6.0-rc3 in the book of main.
  • The landing page, SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-29, for 5.6.0-rc3

  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths and the old error body until 5.6.0 ships.

The review of 2026-10-31, for 5.6.0-rc4

Read on the tree after commit d79ef8b40, for the FerroEHR release candidate 5.6.0-rc4. The review of 2026-10-29 for 5.6.0-rc3 stands for everything that release candidate carried. The facts added since and checked in every text in scope: an AQL lookup by an ELEMENT value reads a value index in both stores (#4437), and a lookup of an EHR by its EHR_STATUS.subject reads the EHR row (#4442); a request on a kept-alive connection idle past the head bound is answered (#4381); the server keeps planner statistics on the partitioned tables and the serving roles hold MAINTAIN on them (#4453), and a version read probes its commit audit by key (#4438); EHR e NOT CONTAINS runs (#4465); a DV_DATE_TIME comparison compares instants (#4466); the REST, party, template and Simplified Format fixes of the Veredictum 0.3.0 catalogue (#4458 to #4464, #4468 to #4475); a token holding only system/ SMART on openEHR scopes passes the role gate on clinical operations (#4467); the viewer draws every page on its server with FerroUI 0.6, with no WebAssembly (#4433); AQL TERMINOLOGY('expand', …) takes the value set from the url argument of a $expand query form (#4488); the conformance pipeline runs the Veredictum 0.3.0 catalogue against one deployment with FerroTERM beside it, in CI through the Veredictum GitHub Action, and keeps no EHRbase comparison (#4448, #4449, #4476); chart 21.0.0-rc4 with appVersion 5.6.0-rc4.

Changed on 2026-10-31, for 5.6.0-rc4

  • The threat model: its grant table listed what each credential reaches and left out the MAINTAIN grant clinical migration 0026 and party migration 0018 give the serving roles; the ferroehr_clinical and ferroehr_party rows name it (#4453).

Read and kept on 2026-10-31, for 5.6.0-rc4

  • The viewer pages: the viewer and security describe the server-drawn viewer, its Content-Security-Policy and its configuration as the image runs them.
  • The operations pages: operations states the MAINTAIN grant and the statistics refresh as the server applies them.
  • The API and query pages: querying with AQL lists NOT CONTAINS as supported and refuses only a read through its variable; security states the system/ scope rule of the role gate.
  • The hazard log and the technical documentation: the logging component is unchanged since 5.6.0-rc3, so neither gains a row.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc4 and image 5.6.0-rc4 in the book of main.
  • The landing page, SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-31, for 5.6.0-rc4

  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths and the old error body until 5.6.0 ships.

The review of 2026-10-31, for 5.6.0-rc5

Read on the tree after commit b3df38e01, for the FerroEHR release candidate 5.6.0-rc5. 5.6.0-rc4 was tagged and not published, so its review above stands for everything it carried. The facts added since and checked in every text in scope: CVE-2026-84243 is judged not_affected in every image (#4500); AQL TERMINOLOGY('validate', …) and TERMINOLOGY('subsumes', …) keep a params_uri value that carries a ? whole (#4490); an AQL query aimed at named EHRs reads their objects first (#4487); the viewer takes its Content-Security-Policy from ferrohealth-server 0.4.2, with the same policy (#4439); the Conformance workflow runs the Veredictum SMART documents (#4450); the storage bench measures on one runner (#4318); the LANG specification is re-vendored and the openehr-* crates step to 0.0.122 (#4377); chart 21.0.0-rc5 with appVersion 5.6.0-rc5.

Read and kept on 2026-10-31, for 5.6.0-rc5

  • The viewer and security pages: the viewer and security state the Content-Security-Policy the viewer sends on ferrohealth-server 0.4.2.
  • The query page: querying with AQL describes TERMINOLOGY() as the engine answers it.
  • The hazard log and the technical documentation: the logging component is unchanged since 5.6.0-rc4, so neither gains a row.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc5 and image 5.6.0-rc5 in the book of main.
  • The landing page, SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-10-31, for 5.6.0-rc5

  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths and the old error body until 5.6.0 ships.

The review of 2026-11-01, for 5.6.0-rc6

Read on the tree after commit df911334d, for the FerroEHR release candidate 5.6.0-rc6. The facts added since 5.6.0-rc5 and checked in every text in scope: the CONTRIBUTION commit reads the body through the released NewContribution schema and refuses the shapes it does not define with 400 (#3595, #4543); the EHDS Art. 3(3) delay of the person’s own access, off by default under [privacy.access_delay] (#4351); the EHDS Art. 5 mark of what a person inserts and the 403 on their change of a professional’s version (#4350); the ADL 2 validation fixes (#4505, #4513, #4522, #4529, #4536, #4540, #4567); the FLAT and STRUCTURED reads under an ADL 2 template (#4551, #4559); the node header columns and the smaller storage per version (#4525); the audit search tokens taken from the record’s columns (#4524); the 400 for a malformed versioned read (#4541); the viewer’s template releases and its unavailable count (#4506, #4533); the conformance pipeline on Veredictum 0.3.2 with the released and development profiles and their README badges (#4503); and the openehr-* crates versioned at publish (#4555); chart 21.0.0-rc6 with appVersion 5.6.0-rc6.

Read and kept on 2026-11-01, for 5.6.0-rc6

  • The README and the conformance page: the badges are drawn from the runs on main under each profile, and conformance states that the released profile alone backs a released claim and that the development profile is never counted toward it.
  • The security, authentication and audit pages: security and the authentication configuration describe the Art. 3(3) delay as a Member State option, off by default, that withholds data from the person and never from a professional, and the audit trail names the feeds that carry records about a held composition as controller feeds.
  • The resources page: resources states the CONTRIBUTION body as the released NewContribution schema reads it.
  • The operations page: the storage sizing section names 5.6.0-rc5 as the measured release and states the node header columns’ effect on it.
  • The hazard log and the technical documentation: the logging component changed since 5.6.0-rc5 (ehds-logging and ihe-atna 0.3.3, the held composition’s records, audit migration 0020), so the hazard log gains review 14, H6 names the hold and its tests, and the technical documentation gains its 5.6.0-rc6 row.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc6 and image 5.6.0-rc6 in the book of main.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-01, for 5.6.0-rc6

  • The conformance run on main had known red rows under the released profile; two of them were Veredictum’s (Vernum-Projecten/Veredictum#964 and Vernum-Projecten/Veredictum#965). The triage on #4455 attributed every row, and the Veredictum 0.3.3 released-profile run on main has none left.
  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths and the old error body until 5.6.0 ships.

The review of 2026-11-01, for 5.6.0-rc7

Read on the tree after commit 3e519c06b, for the FerroEHR release candidate 5.6.0-rc7. The facts added since 5.6.0-rc6 and checked in every text in scope: the EHDS Art. 10 opt-out from the access services, off until [authz.rbac] access_service_roles declares their roles, with the Art. 10(2) vital-interest override (#4516); every write’s committer checked against the authenticated caller and a Bearer caller’s default committer issued by the token issuer (#4520, #4585); the ADL 2 FLAT and STRUCTURED writes, Web Template and validation fixes (#4552, #4559, #4560, #4568, #4574, #4579, #4580); the AQL archetype predicate that matches the namespace exactly and the version as the oldest release (#4629, #4630); the CONTRIBUTION merge inputs and the import of a malformed one (#4577, #4602); the syslog form of an access-delay record naming the held composition (#4569); the viewer’s upload refusals (#4523, #4596); storage generation 4 and the storage bench on one runner (#4525); ferroehr report naming the published openehr-* crate versions (#4565); the response texts without the Service Model vocabulary (#4531, #4634); chart 21.0.0-rc7 with appVersion 5.6.0-rc7.

Read and kept on 2026-11-01, for 5.6.0-rc7

  • The security, authentication and administration pages: security, the authentication configuration and the administration APIs describe the Art. 10 opt-out as a Member State option that FerroEHR ships the software part of, off by default, which hides the EHR from the declared access services only and deletes nothing.
  • The audit page: the audit trail states that the syslog and FHIR forms of the hold’s record name the same access-delay:<vo>.
  • The resources and query pages: resources states the committer check and the CONTRIBUTION merge inputs, and querying with AQL states the namespace and version rule of an archetype predicate.
  • The performance and operations pages: performance describes the storage bench as two builds on one runner and names its bounds, and operations and the command line say which openehr-* versions ferroehr report names, read through openehr_crates_origin.
  • The hazard log and the technical documentation: the logging component changed since 5.6.0-rc6 (the syslog object of an EHR-level record, the opt-out’s access records and the override’s mark), so the hazard log gains review 15, H2 and H9 name their new tests, and the technical documentation gains its 5.6.0-rc7 row.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc7 and image 5.6.0-rc7 in the book of main.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-01, for 5.6.0-rc7

  • The conformance run on main under the released profile fails no row against the server and has one errored row, Veredictum’s own (Vernum-Projecten/Veredictum#964, closed upstream after the pinned 0.3.2, which still errors on it); the triage on #4455 stays open. The badges show the run as it stands, and no text claims a green record.
  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths and the old error body until 5.6.0 ships.

The review of 2026-11-02, for 5.6.0-rc8

Read on the tree after commit 97265816d, for the FerroEHR release candidate 5.6.0-rc8. The facts added since 5.6.0-rc7 and checked in every text in scope: the published Veredictum 0.4.0 record of a released-profile run against FerroEHR on deployment.spec_profile = "stable" with no failed and no errored row, the Veredictum 0.1.6 documents removed, and the released-profile verdict required in the merge queue and the release (#4451, #4455, #4771); conformance_profile derived from that record (#4672); the read-only ADL 1.4 to ADL 2 conversion routes (#4655); a template upload that answers 400 for every invalid template and never refuses a stored one (#4643, #4647); the ITS-JSON minItems of folders, events and languages (#4644, #4648); the canonical XML items root of a LOCATABLE (#4645); an AQL parent archetype that admits a child release only through that release’s own parent release (#4637); a subject resolved through the EHR subject columns and the empty linkage domain removed (#4675, #4684, #4697); the processing-restricted type on a refused write and the curated 500 texts (#4657); the identifier scan reading a DV_IDENTIFIER’s declared type (#4702); the validation refusal’s Prefer advice (#4597); the inline primitive’s node id (#4739); the viewer’s session gate on every unsafe method and LEPTOS_SITE_ADDR no longer read (#4724, #4725); every item-tag operation checked by ABAC in its target’s mode (#4768); reverse_relationships in a stable-profile canonical-JSON party read (#4775); chart 21.0.0-rc8 with appVersion 5.6.0-rc8.

Changed on 2026-11-02, for 5.6.0-rc8

  • Privacy configuration: the identifier scanner was said to refuse a national-identifier value “wherever it sits, DV_IDENTIFIER.id included”. Since #4702 a DV_IDENTIFIER whose type declares an item is not scanned, and one in a FEEDER_AUDIT only when its type names a person, which the same page’s table states. The sentence now points at that table.
  • Kubernetes installation: the production deployment probe was said to run against four PostgreSQL clusters. With the linkage domain gone (#4684) it runs against three, and the text says so.
  • Records of processing: the audit schema was said to sit outside “all three” pseudonymisation domains; since #4684 there are two, the clinical and the demographic.
  • Conformance: the page said the instrument judges FerroEHR on its default development profile. Since #4775 the released run judges it on deployment.spec_profile = "stable", as the workflow sets; the page now names the profile of each run.

Read and kept on 2026-11-02, for 5.6.0-rc8

  • The conformance pages: conformance, the landing page and the README show the Veredictum 0.4.0 badges and the record’s counts per part from the committed record, and state that the released run decides the job. No text types a number the record does not carry.
  • The API pages: using the API derives conformance_profile from the record and names no configuration key for it; problem types states the 400 of a template upload and the processing-restricted type; templates and validation describes the conversion routes as FerroEHR’s own design with their 404 and 422; content negotiation names the root elements no ITS-XML schema publishes.
  • The security and demographics pages: security states that an item-tag operation takes its target’s ABAC mode (#4768), and demographics states that a stable-profile canonical-JSON party read carries reverse_relationships and the canonical-XML read does not yet.
  • The storage and configuration pages: the server configuration refuses [storage.linkage] as an unknown key, operations carries the upgrade past the linkage domain, and the audit trail keeps the linkage access-record domain for a crossing, which the code still writes.
  • The hazard log and the technical documentation: the logging component changed since 5.6.0-rc7 (the operation table, the conversion routes’ access records, the service error type, the batch insert, audit migration 0021, the tag read’s object and the records of the parties a party read discloses), so the hazard log gains review 16 and the technical documentation its 5.6.0-rc8 row; no control changed and every named test exists.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc8 and image 5.6.0-rc8 in the book of main.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-02, for 5.6.0-rc8

  • The development-profile run, shown beside the record and never counted toward it, fails the row tracked on #5258; its badges show the run as it stands, and no text claims it green. The two rows #4776 tracked pass: under the development profile an ADL 2 description without an author or a lifecycle state is refused.
  • A stable-profile canonical-XML party read does not carry reverse_relationships yet (#4778); the demographics page says so.
  • The public sandbox runs 5.5.0 and is not redeployed for a pre-release, so it serves the old paths and the old error body until 5.6.0 ships.
  • The performance page keeps the figures the Veredictum 0.1 line measured until a perf-result record measured on the reference host replaces them, as the page says.

The review of 2026-11-02, for 5.6.0-rc9

Read on the tree after commit e6100379c, for the FerroEHR release candidate 5.6.0-rc9. 5.6.0-rc8 was tagged and not published: its vulnerability record refused findings that had no OpenVEX judgement, so every claim of the 5.6.0-rc8 review above reaches users first with this release. The facts added since 5.6.0-rc8 and checked in every text in scope: the OpenVEX judgements of the gcc-14 and SQLite findings (#4789); the vulnerability record run on a release pull request and every day, and the published images re-scanned every day (#4795); audit migration 0022 and the three unused horizon reap functions removed (#4770); the AOM2 node id of a primitive constraint inside a rule or a slot assertion (#4770); chart 21.0.0-rc8 with appVersion 5.6.0-rc9.

Read and kept on 2026-11-02, for 5.6.0-rc9

  • The supply-chain pages: verifying releases, the supply-chain hardening page and the CRA risk assessment state the daily re-scan and the vulnerability record each release publishes, as the workflows run them.
  • The audit and retention pages: the audit trail and the hazard log’s H5 describe reaping through each record’s own retention end, and no text names a removed reap function; the hazard log gains review 17 and the technical documentation its 5.6.0-rc9 row.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc8 and image 5.6.0-rc9 in the book of main.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-02, for 5.6.0-rc9

  • Everything the 5.6.0-rc8 review left open: the development-profile run’s failed rows (#4776, #4777), the canonical-XML reverse_relationships (#4778), the 5.5.0 sandbox and the performance page’s earlier figures.

The review of 2026-11-02, for 5.6.0-rc10

Read on the tree after commit 797ceb527, for the FerroEHR release candidate 5.6.0-rc10. 5.6.0-rc8 and 5.6.0-rc9 were tagged and not published, so the claims of their reviews above reach users first with this release. The facts added since 5.6.0-rc9 and checked in every text in scope: the outbox served to its named readers over a cursor feed under the __ferroehr.outbox scope, and FerroEHR’s own broker publisher, [events], the subscription routes and the viewer’s Subscriptions page removed (#4786, #4766); the access record of a VERSION read naming the content it served (#4660); the BASE 1.3.0 description invariants of an ADL 2 template under the development profile (#4776); a slow database connect no longer recorded as an outage, and a reset one recorded as an outage on the pool’s own connect too (#4782, #4807); Prefer read on every route as RFC 7240 §2 defines it (#4722); chart 21.0.0-rc9 with appVersion 5.6.0-rc10.

Changed on 2026-11-02, for 5.6.0-rc10

  • The README, the landing page, beyond core and change events said FerroBRIDGE reads the feed and delivers the events to an AMQP broker, and the landing page listed “RabbitMQ via FerroBRIDGE”. FerroBRIDGE’s broker delivery is still in progress (FerroHEALTH/FerroBRIDGE issue 1367), so each text now says so, and the change-events page says a deployment that needs broker events keeps a release with FerroEHR’s own publisher until a FerroBRIDGE release carries it. The removal fragment of the changelog says the same.

Read and kept on 2026-11-02, for 5.6.0-rc10

  • The configuration, integrations, administration API, audit and privacy pages describe [record.change_feed] readers, the feed’s scope and its access records, and the change-events page the removed [events], secrets.eventsUrl and plaintext_broker, which the server refuses at boot.
  • The hazard log and the technical documentation: the logging component’s modules did not change since 5.6.0-rc9; the hazard log gains review 18 and the technical documentation its 5.6.0-rc10 row for the VERSION read’s record and the feed’s access records.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0-rc9 and image 5.6.0-rc10 in the book of main.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-02, for 5.6.0-rc10

  • Broker delivery waits on a FerroBRIDGE release (FerroHEALTH/FerroBRIDGE issue 1367), as the change-events page says.
  • Everything the 5.6.0-rc8 review left open: the development-profile run’s failed row tracked on #5258, the canonical-XML reverse_relationships (#4778), the 5.5.0 sandbox and the performance page’s earlier figures.

The review of 2026-11-04, for 5.6.0

Read on the tree after commit 25db5c69c, for FerroEHR 5.6.0, the stable release after 5.5.0 and the ten release candidates whose reviews stand above. The facts added since 5.6.0-rc10 and checked in every text in scope: an EHR whose access settings do not parse held closed, such settings refused on every write and load, and the administrator’s repair route (#4846, #4854, #4865, #4853); the manifest export checked against the EHR_ACCESS settings of every EHR it names, both EHR Extract export and import routes behind the ABAC pre-check, and an Art. 10(2) override export recorded (#4864, #4869, #4872, #4870); the viewer’s sign-in through ferrohealth-session, its absolute session lifetime and session watch, and its CDR requests confined to the configured CDR (#4873, #3057, #4223, #4954); a version equal to its predecessor stored once, and [record.versioning] record.versioning.idempotent_resend (#4415); storage generation 5 (#4904); the family pins, FerroTERM 0.4.1, FerroPIX 0.0.15 and FerroBRIDGE 0.0.19 by digest, and the newest family crates (#4933, #4981, #4968); the conformance record of Veredictum 0.4.1 (#4971); chart 21.0.0 with appVersion 5.6.0.

Changed on 2026-11-04, for 5.6.0

  • The README, the landing page, beyond core and change events said FerroBRIDGE’s broker delivery was in progress and that a deployment needing broker events keeps 5.5.0 or 5.6.0-rc7. FerroBRIDGE 0.0.15 delivers the outbox feed to an AMQP broker under its [events] lane (FerroHEALTH/FerroBRIDGE issue 1367, closed 2026-11-02), so each text now names that release, and the migration steps on the change-events page apply to FerroBRIDGE 0.0.15 or later.

Read and kept on 2026-11-04, for 5.6.0

  • The chart, its README and the Kubernetes installation say what FerroBRIDGE 0.0.17 checks through [auth], which #4933 corrected where they said FerroBRIDGE authenticated no caller.
  • The configuration and operations pages name the refusal of the OTLP TLS files beside an endpoint that is not https://, the span named by method and route and http_server_active_requests (#4968).
  • The hazard log, the technical documentation and the CRA risk assessment: every test the hazard log names exists; it gains review 19, the CRA risk assessment revision 25, and the technical documentation keeps its 5.6.0 row.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.0 and image 5.6.0.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-04, for 5.6.0

  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for the next release.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-05, for 5.6.1

Read on the tree after commit 75262e1ce, for FerroEHR 5.6.1, the patch release after 5.6.0. The facts added since 5.6.0 and checked in every text in scope: the hosted sandbox’s viewer posture with the session keys the 5.6.0 viewer reads (#4985); the family pins, FerroPIX 0.0.16 with its chart 0.3.8, FerroBRIDGE 0.0.20 and ferroehr-postgres 5.6.0 on the hosted database box by digest, with FerroTERM still at 0.4.1, and four FerroSYS crate releases that change no behaviour (#4987); chart 21.0.1 with appVersion 5.6.1.

Changed on 2026-11-05, for 5.6.1

  • The Kubernetes installation said the chart pins FerroPIX 0.0.15 by digest. #4987 moved the pin to 0.0.16 with the FerroPIX chart 0.3.8, so the page now names 0.0.16.

Read and kept on 2026-11-05, for 5.6.1

  • The README, beyond core and change events name FerroBRIDGE 0.0.15 or later for broker delivery, which 0.0.20 still delivers.
  • Audit and the technical documentation name FerroBRIDGE’s recording of its FHIR facade’s accesses from 0.0.16, unchanged in 0.0.20.
  • The hazard log, the technical documentation and the CRA risk assessment: every test the hazard log names exists; it gains review 20, the CRA risk assessment revision 26, and the technical documentation keeps its 5.6.0 row, since no module of the logging component changed.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.1 and image 5.6.1.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-05, for 5.6.1

  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.7.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-05, for 5.6.2

Read on the tree after commit ea1190926, for FerroEHR 5.6.2, the security-only patch release after 5.6.0; the v5.6.1 tag exists and its release never published. The fact added since 5.6.1 and checked in every text in scope: CVE-2026-76844 in the zlib package of the three images, judged not_affected in security/vex/ferroehr-os.openvex.json and security/vex/postgres-os.openvex.json (#4993); chart 21.0.2 with appVersion 5.6.2.

Read and kept on 2026-11-05, for 5.6.2

  • Verifying releases describes the two OS documents by the packages they cover, zlib among them, and counts six affected findings in ferroehr-postgres; the new statements are not_affected, so both hold.
  • The CRA risk assessment gains revision 27 for the judgement; the hazard log and the technical documentation keep their rows, since no module of the logging component changed.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.2 and image 5.6.2.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-05, for 5.6.2

  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.7.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-05, for 5.6.3

Read on the tree after commit 6b2ff1644, for FerroEHR 5.6.3, the first published release since 5.6.0; the v5.6.1 and v5.6.2 tags exist and neither release published, so this review also covers the facts of the 5.6.1 and 5.6.2 reviews. The facts added since 5.6.2 and checked in every text in scope: the performance guard’s baseline, the newest published release (#4998); RUSTSEC-2026-0440 judged not_affected (#4996); ferrohealth-server 0.5.0 (#5000); FerroBRIDGE 0.0.21 in the chart by digest (#5017); chart 21.0.4 with appVersion 5.6.3.

Read and kept on 2026-11-05, for 5.6.3

  • Verifying releases describes the Rust document as the advisories the gate accepts plus the one a lock-file scanner reports, which holds with RUSTSEC-2026-0440 among the accepted.
  • The README, beyond core, change events and audit name FerroBRIDGE 0.0.15 or later for broker delivery and its facade’s access records from 0.0.16, which 0.0.21 still provides.
  • The CRA risk assessment gains revision 28; the hazard log and the technical documentation keep their rows, since no module of the logging component changed.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 21.0.4 and image 5.6.3.
  • SECURITY.md, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-05, for 5.6.3

  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.7.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-10, for 5.7.0-rc1

Read on the tree after commit a6ce1927d, for the FerroEHR release candidate 5.7.0-rc1. The facts of this release checked in every text in scope:

  • Party binding: every person is bound to a demographic PERSON. A login no party carries is refused with 403 and refused-unbound, and a bound caller commits under the committer the server derives (#5003, #5004, #5005, #5006). The Basic user list holds persons only (#5009).
  • SCIM 2.0 provisioning at /scim/v2/, off by default, with the Entra ID profile and contact erasure (#5014, #5015, #5178). Just-in-time creation of a provisional person is off by default (#5016).
  • The caller kinds: system callers, federation agreements with the Wabvpz Art. 15f roles, the MyHealth@EU contact point route and its caller rate, and representative confinement (#5086, #5100, #5101, #5102, #5107, #5149).
  • The access record carries the binding outcome, and every refused request is recorded (#5118, #5119).
  • The security fixes GHSA-jhg2-jwjp-hq74 and GHSA-pfgx-xq28-2629 (#5003, #5047), the zlib fixes in the three images (#5137, #5152), and the four krb5 CVEs of the PostgreSQL image judged not_affected (#5344).
  • The configuration tree is one table per concern, with every retired key refused by name (#5259). The four outbound integrations use one endpoint table (#5331).
  • The family run path (#5321).
  • Chart 25.0.1-rc1 with appVersion 5.7.0-rc1.

Changed on 2026-11-10, for 5.7.0-rc1

  • The chart description and README said config.deployment_profile names the posture, and the README’s install command set that key, which the chart refuses to render since #5259. #5341 renamed both to config.deployment.profile before this review. It also made the chart’s boot check judge the production overlay’s posture again.
  • The CRA risk assessment’s outbound data flows named [auth] mode = "introspection" with [auth.introspection.client] and [pixm] subject_index. They now name mode = "introspection" in an [[auth.issuer]] table, [auth.issuer.introspection.client] and [record] subject_index (#5259). #5331 had already moved the integrations’ oauth tables to [<x>.endpoint.credentials.oauth]. Revision 29 records the release.
  • The hazard log:
    • H2 named two assurance tests #5003 renamed.
    • H5 named its floor tests in config/mod.rs and system_log/config.rs, which moved to config/audit_tests.rs and system_log/retention.rs.
    • Review 21 records both moves.
  • The technical documentation named the FerroAUDIT pins as 0.3.3. It names 0.9.1, the version the root Cargo.toml pins, and gains the 5.7.0-rc1 row.

Read and kept on 2026-11-10, for 5.7.0-rc1

  • The configuration pages: configuration, the integrations, audit, auth and CLI pages, security, the terminology page and the go-live checklist name the 5.7.0 keys. docs-claims resolves every documented key against the shipped default configuration.
  • The identity pages: identity providers says the Basic list holds persons only and machines use client credentials. The bootstrap and migration page names the four steps and the go-live prerequisites: FerroBRIDGE v0.0.23 and FerroFED v0.0.17 before a deployment adds persons issuers its bridge or gateway writes through.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 25.0.1-rc1 and image 5.7.0-rc1 in the book of main. The frozen book of 5.6.3 under /docs/latest/ keeps its pins, since a pre-release freezes no book.
  • SECURITY.md names security@cadasto.com for a vulnerability report (#5272).
  • The landing page, the crate READMEs, the information sheet, the instructions for use and the image labels make no claim this release changes.

Left open on 2026-11-10, for 5.7.0-rc1

  • The public sandbox runs 5.6.3 until the owner moves it to 5.7.0-rc1 for the reseed #5009 asks of the candidate. The release legs move it to 5.7.0 at the stable cut.
  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258, now for 5.9.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-10, for 5.7.0-rc2

Read on the tree after commit 936d1d921, for the FerroEHR release candidate 5.7.0-rc2. 5.7.0-rc1 was tagged at dd30c84c4 and not published. The review of 2026-11-10 for 5.7.0-rc1 stands for everything 5.7.0-rc1 carried. The facts added since, checked in every text in scope:

  • The release’s performance guard and logging evaluation start the login provisioner the quickstart ships, so they measure the candidate with its Basic login bound (#5346). No product code changed.
  • Chart 25.0.1-rc2 carries appVersion 5.7.0-rc2.

Read and kept on 2026-11-10, for 5.7.0-rc2

  • The logging performance page describes the method and the claims scripts/perf/audit.sh measures. The script now binds its login as the quickstart does, which changes neither.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 25.0.1-rc2 and image 5.7.0-rc2 in the book of main.
  • The landing page, SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-10, for 5.7.0-rc2

  • The public sandbox runs 5.6.3 until the owner moves it to 5.7.0-rc2 for the reseed #5009 asks of the candidate. The release legs move it to 5.7.0 at the stable cut.
  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.9.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-11, for 5.7.0-rc3

Read on the tree after commit 1b79e7f5c, for the FerroEHR release candidate 5.7.0-rc3. The reviews of 2026-11-10 for 5.7.0-rc1 and 5.7.0-rc2 stand for everything those candidates carried. The facts added since, checked in every text in scope:

  • The configuration is redacted by its secret types alone, and every URL and issuer renders with its credentials masked (#5322). Administration APIs says so and names the URL keys.
  • The annotated default configuration is generated from the configuration types, and the reference page includes it (#5316).
  • The family crates move to their newest releases, and every outbound client is built from the family endpoint table (#5320, #5324, #5325). The quickstart, the hosted sandbox and the chart pin FerroPIX 0.0.26 and FerroTERM 0.4.5 (#5351).
  • The audit record takes the purpose of use and the professional role from an IHE IUA token (#5323) and records a contact point’s request message id (#5011).
  • The crates page says the openehr-* crates follow Cargo SemVer, judged by cargo-semver-checks, and that their next publish moves the line to 0.1.0 (#5353).
  • The quickstart compose file attached to a release carries a checksum.
  • Chart 25.0.1-rc4 carries appVersion 5.7.0-rc3.

Read and kept on 2026-11-11, for 5.7.0-rc3

  • The configuration pages: configuration includes the generated reference, which config-reference.sh holds to the types, and docs-claims resolves every documented key.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 25.0.1-rc4 and image 5.7.0-rc3 in the book of main. The Kubernetes installation names FerroPIX 0.0.26.
  • The landing page and the README were aligned with the FerroPIX site in #5382 and make no claim this release changes beyond it.
  • SECURITY.md, the crate READMEs, the information sheet, the instructions for use, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-11, for 5.7.0-rc3

  • The public sandbox runs 5.6.3 until the owner moves it to 5.7.0-rc3 for the reseed #5009 asks of the candidate. The release legs move it to 5.7.0 at the stable cut.
  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.9.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-11, for 5.7.0-rc4

Read on the tree after commit 43bb17040, for the FerroEHR release candidate 5.7.0-rc4. The reviews for 5.7.0-rc1 to 5.7.0-rc3 stand for everything those candidates carried. The facts added since, checked in every text in scope:

  • The hosted sandbox’s configuration declares its SCIM service provider under [auth.scim], and CI reads every server and viewer configuration the repository ships through the program built from the change (#5396). The hosted sandbox README and the bootstrap and migration page name [[auth.scim.client]].
  • The viewer’s OpenID Provider issuer and redirect URIs render with their credentials masked, through ferrohealth-session 0.13.0 (#5394).
  • Chart 25.0.1-rc5 carries appVersion 5.7.0-rc4.

Read and kept on 2026-11-11, for 5.7.0-rc4

Left open on 2026-11-11, for 5.7.0-rc4

  • The public sandbox runs 5.6.3 until the owner moves it to 5.7.0-rc4 for the reseed #5009 asks of the candidate. The release legs move it to 5.7.0 at the stable cut.
  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.9.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-11, for 5.7.0

Read on the tree after commit 52733869b, for the FerroEHR release 5.7.0. The reviews for 5.7.0-rc1 to 5.7.0-rc4 stand for everything those candidates carried. The facts added since, checked in every text in scope:

  • A PERSON binds a login only while it performs a professional ROLE of the party mapping’s role template, and ferroehr party-binding adopt gives an existing PERSON its ROLE (#5004). The bootstrap and migration page says so and orders adopt before the first SCIM run.
  • The hosted sandbox’s reseed brings the CDR up through the deploy script’s deploy verb and names its EHR subjects by FerroPIX pseudonym (#5406, #5409).
  • Chart 25.0.1 carries appVersion 5.7.0.

Changed on 2026-11-11, for 5.7.0

  • The bootstrap and migration page named FerroFED v0.0.17 as the release whose gateway reaches the CDR under the caller, and said it would follow this release. FerroFED v0.0.17 was tagged without that change, FerroFED v0.0.19 does not carry it either, and FerroHEALTH/FerroFED#1615, which carries it, is open in the FerroFED milestone v0.0.20. The page now says no FerroFED release up to v0.0.19 reaches the CDR under the caller and names the issue, and the 5.7.0 changelog section says the same.

Read and kept on 2026-11-11, for 5.7.0

Left open on 2026-11-11, for 5.7.0

  • The public sandbox follows the latest image tags, which the release legs move to 5.7.0 once its images are scanned and tagged.
  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.8.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.

The review of 2026-11-13, for 5.7.1-rc1

Read on the tree after commit de48ac5a8, for the FerroEHR release candidate 5.7.1-rc1. The reviews for 5.7.0-rc1 to 5.7.0 stand for everything 5.7.0 carried. The facts added since, checked in every text in scope:

  • Operational policy lives in the CDR (#5433): who may do what is coded content of demographic parties, read on every request with no restart, under [authz.grants] allowed. System clients with their imports and exports (#5432, #5436, #5530), the backend and DEMOGRAPHIC clients and the clients a token is classified through (#5437), Basic users’ roles and scopes, with [authz.rbac] first_admin as the first administrator (#5438), trusted import systems, origins, change-feed readers and SCIM clients (#5439), federation agreements (#5440), just-in-time admission (#5441) and the four privileges beyond clinical access (#5521) are grants; the audit category map, the access delay and the Cedar policy set are admin registers (#5442). A policy party is recognised by the codes it holds, in any archetype, and only an administrator reads or writes one (#5434, #5450, #5474, #5485).
  • The configuration lists 855 keys, 920 in 5.7.0, and names the 10 a deployment must set (#5455); one public URL takes the place of three (#5465), and the party mapping defaults to the shipped demographic definitions (#5428, #5462).
  • Representatives in the patient portal read the patient’s record under a representation the deployment or the patient records, and a former representative is never taken for the patient (#5446, #5532).
  • The demographic write rules refuse a relationship sourced elsewhere and a reference of the wrong kind, and reverse_relationships reads as it stood at a version (#5471, #5472, #5473).
  • The security fix GHSA-gwh2-7hjm-4pj4: every issuer FerroEHR’s own keys name renders its userinfo masked (#5418).
  • The family crates move to ferrohealth-smart 0.24.1 and its siblings, ihe-atna and ehds-logging 0.18.0 among them (#5417, #5422, #5465, #5437).
  • Chart 32.0.0-rc1 carries appVersion 5.7.1-rc1.

Changed on 2026-11-13, for 5.7.1-rc1

  • The 5.7.1-rc1 changelog section said a policy party is recognised by its archetype, answered 403 naming the archetype, and that no check read a grant yet; a later change of the same release recognises it by its codes and every check reads the grants. It said a Basic user holds its granted roles beside its [[auth.basic.users]] table, which no longer holds roles, and that demographic AQL hides a party by its root archetype. The four entries (#5434, #5435, #5438, #5450) now say what 5.7.1 does.
  • The key counts of the release notes said 868 keys, 921 in 5.7.0. The generated reference lists 855 at this commit, and the 5.7.0 reference listed 920: the 921st key, [auth] described_issuer, came after 5.7.0 (#5425). The #5455 entries, the CRA risk assessment and this review give 855 and 920.
  • The CRA risk assessment: revision 31 and C13 name the client standing, the first administrator, the four privileges #5521 moved, the import and export reach of #5530 and the portal representations, and the crates of #5437. The technical documentation and the hazard log name the FerroAUDIT pins 0.18.0, and both record the ITI-81 subject read and the access-service mark coming from grants (#5521); H6 already names the grant and its test.

Read and kept on 2026-11-13, for 5.7.1-rc1

  • The Operational policy page describes each kind of policy party with a worked example and labels each of its own-design decisions; docs/architecture.md records the model with its citations.
  • The configuration pages: configuration includes the generated reference, and the auth, audit and privacy pages, security and identity providers name the 5.7.1 keys and the grants that replaced the removed ones. docs-claims resolves every documented key.
  • The installation pins: the Kubernetes installation, the chart README and verifying releases name chart 32.0.0-rc1 and image 5.7.1-rc1 in the book of main.
  • The landing page, the README, SECURITY.md, the crate READMEs, the information sheet, the image labels and the chart description make no claim this release changes.

Left open on 2026-11-13, for 5.7.1-rc1

  • The instructions for use at the pinned revision (2026-10-26) name keys 5.7.0 and 5.7.1 refuse at start: deployment_profile, [audit.store] retention_years, [audit.syslog] host, [authz.professional] and [audit.categories], and say FerroEHR ships no category map. The page lives in the FerroHEALTH book, whose repository is paused, so it is tracked on #5534 for the owner’s call.
  • The public sandbox follows the latest image tags and stays on 5.7.0, since a pre-release moves no tag.
  • A v2-namespace OPT 1.4 whose ordinal list keeps xsi:type DV_ORDINAL is accepted, tracked on #5258 for 5.8.0.
  • The performance page keeps its earlier figures until a record measured on the reference host replaces them.