conan-io
Investigate file de-duplication performance for .tgz due to timestamps
The tgz is different for every different compression, due to checked-out timestamps getting into the tgz.
In general this isn't an issue because the revision systems already checks for de-duplication and avoiding transfers of already existing revisions. But if the revision is not fully indexed, this file de-dup might be triggered.
This is one of the related tickets: https://github.com/conan-io/conan/issues/2729
Users requesting to maintain/keep the timestamps of artifacts, not wipe them. There was also there a good analysis in https://github.com/conan-io/conan/issues/2729#issuecomment-1307510530, about what other package managers are doing, and it seems they are keeping the times in the compressed artifacts.
It seems that git doesn't preserve mtimes either, so it would be necessary to completely remove the time information from the tgz for reproducible checksums, but that would go against the other use cases requested.
As it can be seen in this POC in https://github.com/conan-io/conan/pull/19201, that makes fully reproducible tgz by removing both time info and user/group/mode info, renders the .tgz broken (see the CI) for many cases, specially the user/group/mode seems more problematic here. Maybe that part could be omitted.