Citation and archiving

CITATION.cff records the shared op_engine / flepimop2-op_engine software release and preserves the authors and order already attributed in both package metadata files. Maintainers review attribution when contributors/roles change rather than deriving it from commit counts. Core uses MIT; the provider's source grants GPL version 3 or any later version (GPL-3.0-or-later). Record both distribution versions when the provider is used, plus flepimop2/op_system/backend versions relevant to the run.

For published work, cite the exact version and its verified identifier. The initial CFF describes Joshua Macdonald, Carl Pearson, and Timothy Willard, op_engine: Operator-partitioned numerical solvers and flepimop2 provider, version 0.6.0, released 2026-10-09, with the v0.6.0 release URL. Development results also need the commit and any local changes. Unreleased features on main are not retroactively part of 0.6.0; use the exact tag's source and changelog and the documentation version selector.

Run just citation to validate CFF 1.2.0 with pinned cffconvert==2.0.0 in an isolated Python 3.12 tool environment and check authors/version/licenses/date against package metadata and the dated changelog. just ci, release validation and CI include citation checks. Update the CFF version, release date and artifact URL alongside a shared version bump. Add preferred-citation only when a paper's authors, title, venue, date and persistent identifier have been verified; retain the versioned software citation for reproducibility.

Identifier audit

On 2026-10-14, the repository, publication workflows and v0.6.0 release notes contained no software archive DOI, .zenodo.json, archival integration instructions or preferred paper citation. Public DataCite and Zenodo API queries for the exact repository URL returned zero records, and public searches found no verified archive/paper identifier. This does not establish that no maintainer record exists: connected archival accounts and unindexed records must still be checked. No persistent identifier is invented in the citation.

The current workflow publishes GitHub release and package-index artifacts; it does not itself perform an archival deposit. The shared GitHub release URL is the verified identifier until a resolving archive record is confirmed.

ACCIDDA software-paper archive plan

The shared release archive follows the ACCIDDA Zenodo pattern selected for op_system. The archival deposit and resolving DOI still need verification. Finish the JOSS/release blockers before freezing that archival release. The release maintainer should:

1. Check the ACCIDDA Zenodo account for an existing op_engine record/version series and reuse it if present. Avoid duplicate records for the same software.
2. Review attribution and grants; align both package versions, the provider's exact core dependency, dated changelog, citation and release tag. Run just ci, just docs, just qualify and just release-validate.
3. Publish through the existing shared release policy and archive the exact tag for both distributions, released artifacts, and any benchmark fixtures/JSON supporting paper claims. Preserve the commit, checksums, environment/hardware, numerical protocol and versioned documentation links.
4. Verify that the deposit resolves and matches the released artifacts. Record the version DOI in citation metadata, release notes and paper. Distinguish a concept DOI for the version series from the exact release DOI.
5. Add a preferred paper citation when its bibliographic record exists, and revalidate identifiers and CFF syntax before announcing the release.

See Zenodo's GitHub integration and release archiving for account setup or a manual deposit. Readiness PRs do not create a deposit, activate integration, reserve a DOI or publish a release.