SCIO: ScanCode.io SCTK: ScanCode Toolkit
There is too much cruft in detected licenses. We know too much without being to distinguish the forest from the trees. Therefore reporting the primary license detection is important: when we get scan results, we can often get 30 licenses for a single a package and this volume is a problem even if it is correct and it is technically correct.
The goal of this improvement is to:
- Combine multiple related license matches in a single license detection.
- In a license detection, expose a primary license expression in addition to the complete, full license expression.
- Make the logic of selection of the primary license visible, at the minimum with a log of combination and primary license selection operations.
This is for SCTK first.
Status:
This has been completed in SCTK and also included in SCIO. We use an updated --summary option and a new license clarity score for this. We also have LicenseDetections for resources/packages and a top level unique license detections as a summary.
Next steps:
- We can report the declared license and other licenses in the license summary of a full scan. The primary license is based; next is to do the same across each package found nested in a scanned codebase. And also compute an individual license clarity score for each these.
Reporting the set of package files for each package instance is important because it allows for natural grouping of these in one unit.
This has been completed in SCTK and also included in SCIO.
Packages:
- manifest, file-level: package.json
- package: object of its own, and related set of files, not always in the same directory
This is completed in SCTK.
License:
- many detections in a file at different locations, could be merged into a single reported license
- same for primary licenses
This is completed in SCTK.
Copyright:
- Copyright and author detection, which are tracked at the line level
- Holder would be for many copyright detections
This is the same issue as for primary license, but for holders
This has not been completed. This is less critical to complete as the tracing is much simpler and can be done manually in the rare cases where this is needed.
- SCTK: add primary license field in package output and populate this based on package-type/ecosystem conventions.
- SCTK: also populate secondary license fields
- SCIO: add primary license field in DiscoveredPackage models and feed it with the data from packages
- SCIO: Do we track secondary? or is this just data aggregated on the fly.
- SCIO: Refine primary license based on license in "key files"
- This is closely tied to the primary license detection and should focus on package manifests and key files.
- Support copyright parsing from all package ecosystems.
- SCTK: See https://github.com/nexB/scancode-toolkit/projects/10 - Work on the model - Work on updating the package code focus in npm, pypi, maven, go.
- SCIO: adopt the two levels manifests/package instances
- Add top level reference licenses section in the JSON output
- Report detections at the file level, one per matched rule
- Report multiple detections for one or more license expressions in a file, eventually grouping multiple detection in a single expression in a file
- Reported detections of Copyrights and authors detection, tracked at the line level in a file
- Holder would be for many copyright detections
- Finish and merge unknown license detection (this depends on completion of 4. Go to two-level reporting of detections for license)
- Update scancode-analyze to the new two-level reporting of license detections
- Revamp how common list of suprrious licenses are detected (this is a bug)
- Use important key phrases for license detection aboutcode-org#2637
This is mostly completed, for follow up see aboutcode-org#2878