Maintaining relevance in 2025

[hmaarrfk]

Comment:

I'm wondering if there is more we can do to shed maintenance burden in 2025

• Decreasing the test matrix for the test images. We likely don't need to test for Ubuntu 16.04 for example: https://github.com/conda-forge/miniforge/blob/main/build_miniforge.sh#L19
  • Should keep us more nimble
• Removing the Darwin name from the OSX installers: https://github.com/conda-forge/miniforge/blob/main/.github/workflows/ci.yml#L97
  • Should shed about a few installers and mental load from the "downloads page"
• Dropping mamba
  • This is more controversial, but with conda-libmamba-solver the performance of conda has greatly improved.
  • With the recent challenges getting mamba 2 rolled out I can't say I'm too excited to try again.
  • perhaps we have a transition packages where mamba is a pointer to conda but emits a warning when used?

Followup of #602

[jakirkham]

Probably latest and LTS are enough for Ubuntu. Though would be interested to hear from others

Maybe same story with Debian (or is testing on Ubuntu enough?)

CentOS 7 testing makes sense given how much we use it in builds

Know the goal is shedding, but it may make sense to test UBI 8 (effectively a limited RHEL 8 in a container), AlmaLinux 8 and/or Rocky Linux 8 given the current EL landscape. Think we need to test something that is not EOL, but could see wide use

---

One less installer name with the same functionality? Sounds like a good change

---

Am hopefully that after the initial churn mamba 2 becomes more stable and friendly. The Mamba team fixed a lot of issues in that upgrade. Think that should be lauded. Wouldn't want teams to feel reluctant to fix complex issues that need breaking changes because of potential reception