You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Noticed while resolving licences in #43. Of the 909 entries in tools_index.json, a substantial share are not software with a licence, and some are not tools at all.
Three classes are detectable by shape, and together they are a lower bound of 54 entries:
shape
count
examples
a descriptive phrase rather than a name (5+ words)
35
Agilent 1290 Infinity UHPLC system, asari MassGrid class (build_grid_sample_wise, add_sample, ...)
Beyond those, the catalogue holds entries that are shaped like names but are not tools in the sense the collection means: C++, Bioconductor, Docker, .NET Framework, stats, Anaconda. And it holds vendor hardware and proprietary products — Agilent MassHunter, Agilent 6550 iFunnel Q-TOF mass spectrometer — which are real, describable, and worth keeping, but have no repository to resolve a licence from.
Why it matters now
They are all license_tier: unknown, alongside genuinely unresolved software. unknown is a correct statement about each of them, but for different reasons, and lumping them together is the same conflation the tier split fixed on the other axis:
a module path or a bare URL is an extraction defect and should not be a catalogue entry;
an instrument has no software licence, so the tier is not merely unknown but inapplicable;
vendor software is genuinely restricted once identified, and saying so is more useful than unknown;
A count of "713 unknown" therefore overstates how much licence resolution is left to do, and understates how much catalogue cleaning is.
Suggested direction
Drop entries whose name is a module path or a bare URL — these are extraction defects with no downstream value.
Decide whether instruments belong in a tool catalogue. If they do, they need a state that says the licence question does not apply, rather than borrowing unknown.
Identify vendor products and tier them restricted, which is what they are.
Only then is the residual unknown count a measure of outstanding licence work.
Step 2 is a governance decision and belongs with a maintainer; steps 1 and 3 are mechanical once it is made.
Noticed while resolving licences in #43. Of the 909 entries in
tools_index.json, a substantial share are not software with a licence, and some are not tools at all.Three classes are detectable by shape, and together they are a lower bound of 54 entries:
Agilent 1290 Infinity UHPLC system,asari MassGrid class (build_grid_sample_wise, add_sample, ...)massdash.loaders.ResultsLoader,massdash.peakPickers.MRMTransitionGroupPickergithub.com/bittremieux/cosine_neutral_loss,github.com/HassounLab/ESPBeyond those, the catalogue holds entries that are shaped like names but are not tools in the sense the collection means:
C++,Bioconductor,Docker,.NET Framework,stats,Anaconda. And it holds vendor hardware and proprietary products —Agilent MassHunter,Agilent 6550 iFunnel Q-TOF mass spectrometer— which are real, describable, and worth keeping, but have no repository to resolve a licence from.Why it matters now
They are all
license_tier: unknown, alongside genuinely unresolved software.unknownis a correct statement about each of them, but for different reasons, and lumping them together is the same conflation the tier split fixed on the other axis:restrictedonce identified, and saying so is more useful thanunknown;A count of "713 unknown" therefore overstates how much licence resolution is left to do, and understates how much catalogue cleaning is.
Suggested direction
unknown.restricted, which is what they are.unknowncount a measure of outstanding licence work.Step 2 is a governance decision and belongs with a maintainer; steps 1 and 3 are mechanical once it is made.