Skip to content

0.0.24

Choose a tag to compare

@github-actions github-actions released this 10 Nov 03:41
· 52 commits to main since this release
Immutable release. Only release title and notes can be modified.
v0.0.24
b26dcc8

Changed

  • The ITI-21 RSP^K22 is checked against RSP_K21 as IHE ITI TF-2 constrains it before it leaves: QRI optional (Table 3.21-6) and QPD-3 repeating, one parameter a repetition (§3.21.4.1.2.3), through ferrobridge-hl7v2 0.4.0's IHE_ITI_21. An answer that does not conform, such as a PID with no name, is answered AE, as the RSP^K23 is (#538).
  • The compose quickstart and the Helm chart's CI values run the development OpenID Provider mock-oauth2-server 6.1.5, and the release and image scans run Trivy 0.76.2 (#557).
  • ihe-iti 0.0.84 requires ferrohealth-smart 0.15 (caret 0.15.0) and ihe-atna 0.9 (caret 0.9.0), and nl-generic-functions 0.0.51 requires ferrohealth-smart 0.15 and ihe-iti 0.0.84. Both name ferrohealth-smart and ihe-atna types in their public API, so a product on those lines needs this release; no signature of either crate moves. Upgrade: move ferrohealth-smart to 0.15.0 and ihe-atna to 0.9.0 beside them (#561).
  • Breaking for an operator: the server and the console move to ferrohealth-smart 0.15.1, ferrohealth-session 0.9.0, ihe-atna 0.9.1, ferrohealth-config 0.4.0, ferrohealth-server 0.8.0, ferrohealth-telemetry 0.4.0 and ferrohealth-usage-report 0.4.0, so the lock holds one copy of each family crate. Every span of time and size in the sections those crates own is now a string with its unit, under a key that names none, and a bare number is refused. Upgrade: in [server] write body_limit = "1MiB", request_timeout, header_read_timeout, idle_timeout, drain_delay, shutdown_timeout, retry_after and compression_min_size (each was …_ms, …_s or …_bytes) and [server.tls] reload_interval; in each [[auth.issuer]] write clock_skew, jwks_max_age, jwks_refresh_interval and timeout, with introspection.cache_max_age, [auth.basic] cache_ttl and the OAuth tables' refresh_margin and timeout; move auth.ca_file to [auth.tls] trust_roots_file, and the OTLP collector's trust_roots_file and client_identity_file under [telemetry.otlp.tls]. In the console write [session] sign_in_timeout, idle_timeout and absolute_timeout ("30min", not idle_timeout_s = 1800) and [oidc] timeout, and move oidc.ca_file to [oidc.tls] trust_roots_file. Each old key fails the start as an unknown key, and the loader names the new one where it can. The [auth] keys of the shapes before ferrohealth-smart 0.9 now fail as unknown keys too, without naming their replacement (#561).
  • Breaking for an operator: FerroPIX's own sections write every span of time and size the same way, under a key that names no unit. Upgrade: [audit] spool_max_size = "64MiB" and spool_write_timeout, with the delivery's timeout and retry.max_backoff under [audit.endpoint] (below); [outbox] retention = "30d"; [pmir.feed] initial_backoff and max_backoff; [hl7v2] frame_limit, idle_timeout and frame_timeout, each consumer's connect_timeout and reply_timeout, and [hl7v2.notification] initial_backoff and max_backoff; [xcpd] delivery_initial_backoff, delivery_max_backoff and delivery_timeout; the terminology server's timeout with its retry.initial_backoff and retry.max_backoff under [terminology.endpoint] (below); and the console's [pix] timeout. In this release each old key refuses the start and names the key that replaces it, all of them in one refusal; from the next release each fails as an unknown key (#561).
  • Breaking for a crate user: ihe-iti 0.1.0 requires ferrohealth-smart 0.16 (caret 0.16.0) and ihe-atna 0.10 (caret 0.10.0), and nl-generic-functions 0.1.0 requires ferrohealth-smart 0.16 and ihe-iti 0.1.0. Both name ferrohealth-smart and ihe-atna types in their public API, so the move of those public dependencies takes each crate's minor step, and a product on those lines needs this release; no signature of either crate moves. Upgrade: move ferrohealth-smart to 0.16.0 and ihe-atna to 0.10.0 beside them (#571).
  • Breaking for an operator: the server reaches the audit repository and the terminology server through ferrohealth-smart's outbound client, each configured by its endpoint table, so the lock holds one copy of each family crate (ferrohealth-smart 0.16.2, ferrohealth-session 0.10.0, ihe-atna 0.10.0, ferrohealth-config 0.4.1, ferrohealth-server 0.8.1, ferrobridge-term 0.3.0, ferrobridge-hl7v2 0.5.0). Upgrade: write the audit repository as [audit.endpoint] with url, timeout and [audit.endpoint.retry] max_backoff (was audit.url, audit.timeout_ms and audit.retry_max_ms); its max_backoff is also the wait between two delivery rounds, and the table takes connect_timeout, retry.max_attempts, tls and credentials (HTTP Basic or the client credentials grant) as well. Write the terminology server as [terminology.endpoint] with url, timeout, retry, credentials and tls (was terminology.base_url, timeout_ms, [terminology.retry], [terminology.credentials] and [terminology.tls]); compose and the Helm chart (0.3.16) name it FERROPIX__TERMINOLOGY__ENDPOINT__URL and config.terminology.endpoint.url. In this release each old key refuses the start and names its place under the endpoint table; from the next release each fails as an unknown key (#571).
  • A configuration the server or the console does not start on is refused with a line that opens cannot start (exit 78), and healthcheck with one that opens not up (exit 1), through ferrohealth-server's run path; the server's line opened the configuration is refused before (#571).
  • ihe-iti 0.1.0 and nl-generic-functions 0.1.0 require the FerroBRIDGE crates by caret ranges, fhir-search 0.1.130 and fhir-types 0.1.130, where ihe-iti 0.0.84 required fhir-search exactly, so a product whose other family crates take another compatible patch of them builds with one copy (#543) (#571).
  • The keys of the family sections that v0.0.24 renamed are refused by name in this release, each with the key that replaces it, all of them in one refusal, read from the file, the environment and --set before the rest of the configuration. On the server: [server] request_timeout_ms, header_read_timeout_ms, idle_timeout_ms, body_limit_bytes, drain_delay_ms, shutdown_timeout_ms, retry_after_s, compression_min_bytes and tls.reload_interval_s; auth.ca_file, auth.basic.cache_ttl_seconds, auth.dpop.max_age_seconds and clock_skew_seconds; each [[auth.issuer]] entry's clock_skew_seconds, jwks_max_age_seconds, jwks_refresh_interval_seconds, timeout_ms and introspection.cache_max_age_seconds; the OAuth tables' refresh_margin_ms, timeout_ms, ca_file, client_certificate_file, client_key_file and previous_key_overlap_ms; and the OTLP collector's trust_roots_file and client_identity_file. On the console: [session] sign_in_timeout_s, idle_timeout_s and absolute_timeout_s, [oidc] timeout_s and ca_file, [oidc.tls] client_certificate, client_key and client_key_file, and the OTLP collector's files. From the next release each fails as an unknown key (#575).

Removed

  • The named refusal of [xcpd] client_certificate_file, client_key_file and reply_trust_file, which v0.0.22 retired in favour of [xcpd.tls]. Each key now fails the start as an unknown key of [xcpd] (#561).

Fixed

  • The server exports its spans and metrics over the TLS [telemetry.otlp.tls] names: it installs its telemetry through ferrohealth-server's installer, which reads the collector's trust roots and client identity, where before it built the exporter without them (#571).

Support period

Only the newest release is supported: there are no maintenance branches, no long-term-support line and no backports, and upgrading is free. Under the Cyber Resilience Act, Cadasto B.V. commits to keep publishing security fixes for at least five years after each release (Art 13(8)) and ships them only in the newest release (Art 13(10)). For this release that commitment runs until the end of November 2031 (Regulation (EU) 2024/2847 Art 13(19)); see SECURITY.md.