It'd be nice to reduce the footprint of our decompression stack to only the decompression schemes Python packaging (widely) uses. One that we could potentially eliminate is xz/lzma, since only a tiny minority of Python wheel distributions (should) be using it.
This would technically be a breaking change. However, it should be a marginal one, since PyPI itself doesn't allow any ZIP compression schemes other than DEFLATE and STORE (and eventually zstd):
https://github.com/pypi/warehouse/blob/6ba4eb2cb1482403a64805fc0d1839e4f36d1bb3/warehouse/forklift/legacy.py#L313
It'd be nice to reduce the footprint of our decompression stack to only the decompression schemes Python packaging (widely) uses. One that we could potentially eliminate is xz/lzma, since only a tiny minority of Python wheel distributions (should) be using it.
This would technically be a breaking change. However, it should be a marginal one, since PyPI itself doesn't allow any ZIP compression schemes other than DEFLATE and STORE (and eventually zstd):
https://github.com/pypi/warehouse/blob/6ba4eb2cb1482403a64805fc0d1839e4f36d1bb3/warehouse/forklift/legacy.py#L313