ADR 0028: cc_by_nc, abstract excerpts and author manuscripts: the rules stand until counsel accepts a change

Status: Accepted
Date: 2026-09-16
Deciders: Founder

Context

The founder asked that publisher-copyright papers show “metadata, abstract excerpt and link”. The verified rendering sets (api/app/services/documents.py: SPAN_LICENCES = {cc0, cc_by, cc_by_sa, customer_internal}, QUOTE_ONLY_LICENCES = {cc_by_nd}, LOCATOR_ONLY_LICENCES = {cc_by_nc, publisher_copyright, unknown}) withhold chunk bodies for locator-only codes, abstracts are ingested as their own chunk under abstract_licence_code defaulting to publisher_copyright, and author manuscripts (is_manuscript) are excluded from full text. The critics found a draft that would have rendered cc_by_nd with figures and shown abstract bodies for locator-only sources.

Decision

No licence rendering rule changes in phase 4. cc_by_nd stays quote-only (cited quote in context, metadata, “Open at publisher”; no figure capture, no thumbnail, no PDF viewer, no export of derived text). The quote carries an attribution line (CC BY-ND requires attribution) under the chip class no_derivatives with the fixed text “open access, no derivatives”; this is display beside the quote, not a rendering-set change, and the API and web renderers agree on the flag (contract 12.3, recorded 2026-09-16). cc_by_nc, publisher_copyright and unknown stay metadata and link.
An abstract body is shown only when the abstract chunk’s own licence_code (the chunk_licences override that abstract_licence_code sets at ingest) is in SPAN_LICENCES, or as a quote when it is in QUOTE_ONLY_LICENCES, capped at 400 characters; otherwise the row reads “metadata and link”. licence_statement is display text beside the chip and never decides rendering; the switch stays chunk_body_allowed(licence_code) per code. Test: “abstract body withheld when abstract_licence_code is publisher_copyright”.
Author manuscripts stay excluded from full text; FULL_TEXT_OK = {CC0, CC_BY, CC_BY_SA} and pmc_redistributable are unchanged.
Three options are recorded for counsel and change nothing before acceptance: (1) cc_by_nc full span for non-commercial internal research use; (2) abstract excerpts for publisher-copyright sources under a quotation rationale; (3) author-manuscript full text where the funder mandate permits. Acceptance of any option supersedes this record with a new decision and moves through the renderer-per-licence test.

Consequences

The renderer-per-licence tests (api/tests/test_licence_rendering.py, web/components/console/evidence-panel.test.tsx) pin the switch against the three sets; this decision is the only path to change them.
“Metadata and link” is honest copy for most of the literature a scientist cites; the “Gather” and “Find the source” controls make the open-access path visible.

Alternatives considered

Abstract excerpts for every source: assumes a right the platform has not verified; rejected.
Thumbnails for no-derivatives articles: a resized thumbnail is a derivative; rejected.

Follow-ups

Counsel review of the three options; a dated CLAIMS-REGISTER.md row if any is accepted.