Release Retrospective to correct Apache MXNet source licenses in post MXNet 1.0 release.
A. Problem Description
Links to email threads and github issues that record the existing issues during various voting periods -
- https://github.com/apache/incubator-mxnet/issues/8913
- Email from Henri Yandell
- Email from Justin McClean
- Results of Apache RAT are in this text file - YetToFix.txt (Some of these are now fixed)
B. Apache Licensing Policies
- LICENSE file requirements
- LICENSE requirements for distribution artifacts with multiple licenses
- NOTICE file requirements (Check Copyright year) - see also https://www.apache.org/legal/src-headers.html
- Apache Legal
- Acceptable and Unacceptable Dependency Licenses
C. Open Pull Requests to fix these Issues (Review help requested!)
- PR to fix the top level LICENSE file
- PR to update the license_header.py file and add apache licenses where missing - Merged.
PR to fix License headers in specific files - Part 1 - Merged
PR for some fixes based on Apache RAT failures - Part 2 - Merged
- PR for some fixes based on Apache RAT failures - Part 3 - Merged
- PR for some fixes based on Apache RAT failures - Part 4
D. Open Issues/Questions/Doubts/Concerns (Help Requested!)
CONCERN 1: What all can be excluded from RAT checks? Currently the following are being excluded from the RAT checks. Is this ok ?
Name of File/Folder Excluded from RAT Check | Reason for Ignoring | Concerns | |
---|---|---|---|
1 | These File Types: *.xml ; *.css ; *.txt; *.md ; \..* ; *.ipynb ; *.html ; *.js ; *.json ; *.svg; *.config; *.names; *.csv | ||
2 | Submodules - 3rdparty/* ; dmlc-core/* ; mshadow/* ; dmlc-core/* ; dlpack/* ; nnvm/* ; ps-lite/* | None | |
3 | R-package | Not a part of MXNet Release | None |
4 | Ignore all Dockerfiles - docker/*; Dockerfile* ; docker_multiarch/* | Dockerfiles cant have license | |
5 | perl-package/* | I am not entirely sure about licensing here. (Also see issue 12 in the last table below) | I think I can add ASF header. |
6 | contrib/* | I am not entirely sure about licensing here. | Is ASF header ok? |
7 | __init__.py files | These files contain no text | None |
8 | docs/* | A header might affect website. | Can someone verify its ok to add ASF without impacting the website |
9 | This file - src/operator/nn/pool.h | It was decided that this file should not have an Apache License and it was removed here - PR 9170 | None. But can be verified again. |
10 | This file - src/operator/special_functions-inl.h | It was decided that this file should not have an Apache License and it was removed here - PR 9170 | None. But can be verified again. |
11 | example/rcnn/rcnn/cython/* | This is licensed under MIT but RAT doesnt pick that. ASF should not be added | Should MIT license text be added explicitly? |
12 | This Dataset - example/gluon/tree_lstm/dataset.cPickle | This is a dataset | None. But can be verified again. |
13 | This file - tools/coreml/pip_package/README.rst | This is a README | None. But can be verified again. |
CONCERN 2: There are still 7 files with unknown licenses as per Apache RAT - SEE TABLE BELOW for this LIst
- This list assumes all 7 PRs listed above are approved and merged
This list assumes above excluded folders and formats are acceptable.
S.No File Name Comments 1. CODEOWNERS 2. appveyor.yml 3. readthedocs.yml 4. snap.python 5. snapcraft.yaml 6. python/mxnet/cython/base.pyi 7. tests/ci_build/pylintrc
CONCERN 3: Issue 6 in table below
This file has had some conflict - example/image-classification/predict-cpp/image-classification-predict.cc
(and one more)
CONCERN 4: Issue 23 and Issue 37 in table below
This Folder needs to be reviewed - src/operator/contrib/ctc_include
E. More Details about the Issues & their Status
No | Category | Problem | Source Files | Suggested by | Leads | Comments |
---|---|---|---|---|---|---|
1. | Source tree | * Move the various git submodules into third-party/ or similar so it's simpler to see what is Apache original source when we review a release. | submodules | Henri Yandell | Haibin Lin | |
2. |
|
|
|
| ||
3. |
|
|
| Henri Yandell | Meghna Baijal | PR Merged |
4. | Comments | * Comment added to CODEOWNERS to explain the file so we don't cause community problems | CODEOWNERS | Henri Yandell | Steffen Rochel | |
5. | LICENSE | * There was a suggestion to simplify the LICENSE to not explicitly list which packages are under each license. Something to consider. | LICENSE | Henri Yandell | Meghna Baijal | If this is done, it would resolve points 8-11 and 13-19 of this wiki |
LICENSE | * Update the Paths to license files once submodules are moved | LICENSE | Meghna Baijal | Haibin Lin | ||
Automated Check | * Update the license_header.py script instead of manual exclusions | tools/license_header.py | Meghna Baijal | Meghna Baijal | makefiles, | |
ISSUES IN SPECIFIC FILES | ||||||
6. | Specific Files | * Resolve License Header - if you follow the URL mentioned in the file it is unclear if the code came form that site or was written for the project by the author of that site. | example/image-classification/predict-cpp/image-classification-predict.cc
| Justin McClean | ||
7. | Specific Files |
|
|
| Meghna Baijal | Possibly RAT got confused by the matching string. Does not need a fix. |
8. | Specific Files |
|
|
| Meghna Baijal | Submodule, not to be fixed as part of MXNet release |
9. | Specific Files | Justin McClean | Meghna Baijal | Removed details from LICENSE. Correctly added MIT license header and in LICENSE file in directory | ||
10. | Specific Files | Justin McClean | Meghna Baijal | Removed details from LICENSE. Correctly added MIT license header and in LICENSE file in directory | ||
11. | Specific Files | Justin McClean | Meghna Baijal | Removed details from LICENSE. Correctly added MIT license header and in LICENSE file in directory | ||
12. | Specific Files | * Missing License Header - is it Apache License? | perl-package/AI-NNVMCAPI/Makefile.PL | Justin McClean | ||
13. | Specific Files |
|
| Justin McClean | Add BSD to license. Packages not named in LICENSE anymore | |
14. | Specific Files |
|
| Justin McClean | Add BSD to license. Packages not named in LICENSE anymore | |
15. | Specific Files |
|
| Justin McClean | Meghna Baijal | |
16. | Specific Files | * Add to LICENSE - BSD license | example/ssd/dataset/pycocotools/coco.py | Justin McClean | ||
17. | Specific Files | * Add to LICENSE - 6 files are BSD licensed | example/rcnn/rcnn/pycocotools | Justin McClean | ||
18. | Specific Files | * Add to LICENSE - BSD license | dmlc-core/cmake/Modules/FindCrypto.cmake | Justin McClean | Submodule, no edits Packages not named in LICENSE anymore | |
19. | Specific Files | * Add to LICENSE - BSD license | cub/experimental/spmv_compare.cu | Justin McClean | Submodule, no edits Packages not named in LICENSE anymore | |
20. | Specific Files | * Incorrect License Header - Has ASF header but is it BSD | prepare_mkl.sh | Justin McClean | Does not say BSD, add? | |
21. | Specific Files | * Incorrect License Header - Has ASF header but is it BSD | src/operator/nn/im2col.h | Justin McClean | Created PR, in review | |
22. |
| * Incorrect License Header - Has ASF header but is it BSD |
| Justin McClean | First was fixed for 1.0.0 here - https://github.com/apache/incubator-mxnet/pull/9170 | |
23. | Specific Files | * Incorrect License Header - Has ASF header but is it BSD | src/operator/contrib/ctc_include/contrib/moderngpu/include/mgpuenums.h | Justin McClean | License seems to be correct. No ASF header. | |
24. | Specific Files | * Incorrect License Header - Has ASF header but is it BSD | example/ssd/dataset/pycocotools/coco.py | Justin McClean | Removed ASF header in PR | |
25. | Specific Files | * Incorrect License Header - Has ASF header but is it MIT | example/rcnn/rcnn/cython/setup.py | Justin McClean | Removed ASF Header in PR | |
26. | Specific Files | * Incorrect License Header - Has ASF header but is it MIT | example/rcnn/rcnn/cython/nms_kernel.cu | Justin McClean | Removed ASF Header in PR | |
27. |
|
|
| Justin McClean | Repeat | |
Specific Files | Resolve License: should this file get an apache license? (RAT thinks so but why does the script skip it?) | src/operator/special_functions-inl.h | ||||
APACHE RAT CHECK FAILURES | ||||||
28. | RAT Failure | * Fix Submodules - RAT detected almost 2000 files with unknown licenses in submodules | Submodules (nnvm, dlpack, 3rdparty, ps-lite, mshadow) and R-package | Decision needed on how to handle submodule licenses | ||
29. | RAT Failure | * Check Docs - RAT detected almost 200 files with unknown licenses in the Docs directory | /docs | Excluded for now | ||
30. | RAT Failure | * Fix dockerfiles without license headers | /docker | Meghna Baijal | Added a top level License.md to this folder | |
31. | RAT Failure | * Fix docker_multiarch - unknown license header | /docker_multiarch | Meghna Baijal | Added a top level License.md to this folder | |
32. | RAT Failure | * Fix scala-package - unknown license | /scala-package | Added apache license | ||
33. | RAT Failure | * Fix tools | /tools | |||
34. | RAT Failure | * Fix tests | /tests | All fixed except one `tests/ci_build/pylintrc` | ||
35. | RAT Failure | * Fix examples | /examples | |||
37. | RAT Failure | * Fix ctc_include | src/operator/contrib/ctc_include | |||