mikemccand commented on PR #14178: URL: https://github.com/apache/lucene/pull/14178#issuecomment-2832351612
Thanks @kaivalnp -- the licensing situation with Conda is indeed complex. I found this nice high-level summary of the changes (which happened ~ 1 year ago, March 2024): https://redresscompliance.com/anaconda-licensing-changes-explained The two paths for installing Python packages/stacks (Conda vs pip/PyPi) also causes confusion/cost, e.g. PyTorch community stopped publishing artifacts through Conda, ~7 months (October 2024) after that license change: https://github.com/pytorch/pytorch/issues/138506 -- apparently, the build infra costs were too high given the smallish % of users that installed via Conda vs PyPi. There was a [recent energetic Reddit discussion about this confusing situation as well](https://www.reddit.com/r/Python/comments/1idsbj7/pytorch_deprecatea_official_anaconda_channel/?utm_source=share&utm_medium=web3x&utm_name=web3xcss&utm_term=1&utm_content=share_button). Thanks @kaivalnp -- the licensing situation with Conda is indeed complex. I found this nice high-level summary of the changes (which happened ~ 1 year ago, March 2024): https://redresscompliance.com/anaconda-licensing-changes-explained The two paths for installing Python packages/stacks (Conda vs pip/PyPi) also causes confusion/cost, e.g. PyTorch community stopped publishing artifacts through Conda, ~7 months (October 2024) after that license change: https://github.com/pytorch/pytorch/issues/138506 -- apparently, the build infra costs were too high given the smallish % of users that installed via Conda vs PyPi. Also, publishing artifacts through the two channels had led to scary/confusing things like substantially different runtime performance depending on whether you installed PyTorch via Conda or PyPi, sheesh: https://github.com/pytorch/pytorch/issues/2163 The `conda-forge` (community maintained, and not subjected to Anaconda's confusing commercial license) channel does have Faiss (https://anaconda.org/conda-forge/faiss-cpu), but it's currently at 1.9.0 -- maybe the developers who build Faiss here might push 1.11.0 (just released) soon? I'm not sure what the build process is behind the scenes ... that page says "copied from [cf-staging / faiss-cpu](https://anaconda.org/cf-staging/faiss-cpu)", which seems to be the staging area to validate the artifacts submitted by the community. Maybe we could build Faiss 1.11.0 and submit it! I have no idea what the process is like ... indeed, foreign to this Java developer heh. Oh, it's described in detail here: https://conda-forge.org/docs/maintainer/understanding_conda_forge/life_cycle/ ... maybe if we wait a bit, 1.11.0 will appear in `conda-forge`? Eventual consistency! [A quick summary of where we stand here: Faiss itself is [generously open source licensed (MIT License)](https://github.com/facebookresearch/faiss/blob/main/LICENSE), but its binaries are distributed (exclusively?) through Conda. `miniforge` (one of the tools for installing/managing Anaconda packages, like `miniconda` but with proper defaulting to the community Conda channels maybe?) is also [generously licensed (BSD 3-clause)](https://github.com/conda-forge/miniforge/blob/main/LICENSE) (both its source code, and the one-time-use tools (hmm also MIT and OpenSSL licensed) used during the installation/local building of miniforge in your env), but when running `miniforge` against one of Conda's channels, that channel has its own licensing covering the binary artifacts it provides, and that's our current wrinkle: we are unsure about the license of the PyTorch Conda channel...] -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
