Attribution and licensing

The effective licence is the union over every source the repository holds, including sources no default code path loads — and there unknown is not permissive, so a source naming no terms is carried unresolved rather than rounded down to fine.22The full computed string is non-commercial: yes; share-alike: yes; attribution: required; redistribution: none; SHARE-ALIKE IN FORCE: derivative works must be released under the same licence; 27 source(s) with UNKNOWN licence (adni, ds000113, hcp-young-adult, mne-sample, mne-spm-face, ram-intracranial, things-eeg2, tuh-eeg, ukbiobank-brain-imaging, buckner2011, conte69, desikan2006, destrieux2010, enigma_hcp_sc, fsaverage, glasser2016, goulas_autoradiography, hcps1200_maps, hill2010, julich_brain, margulies2016, markov2014, netneuro_lausanne_sc, neuromaps, raichle_metabolism, schaefer2018, sydnor2021) — unknown is not permissive.33is_vacuous_licence_text exists because a registry entry once read “See repository LICENSE (open, academic use)” for an atlas whose actual licence imposes no academic-use limit at all — an invented restriction, since corrected. The classifier now refuses to resolve text that names no terms. The Hansen receptor maps are CC-BY-NC-SA-4.0, non-commercial and share-alike, and where those terms attach a derivative must carry them too.44Two questions have to be kept apart or the answer misreports: does the object contain Hansen data, and does the default prior read it? Those have different answers today; the audit is reports/licence_audit.md. The Tian subcortical atlas is unrestricted subject to one condition — that any publication using it cites Tian et al. (2020) — which the table below is how we meet.55Verified against the vendored licence text at assets/src/tian_subcortex/license.txt, not against the registry's summary of it. Schaefer 2018 labels are MIT, but the Genomics Superstruct Project data underneath is released “under its own terms” and those terms are not named, so the classifier resolves it to unknown rather than to MIT. Whether a model trained on CC-BY-NC-SA data is itself a derivative of that data is recorded as unsettled -- no answer asserted and deliberately not answered here; the conservative reading, assume it is, is the one that fails safe.scwbd/release/manifest.py And several datasets carry redistribution_class: none — HCP Young Adult, ADNI, UK Biobank, TUH-EEG among them — so this site hosts no copy of any source and links to no derived data, with access going through each provider's own agreement.