Latest events such as a migration to larger VPS instance due to diskspace shortage and re-enabling the link checker, which showed that a lot of disk space and link checker capacity is taken by the old EoLed repositories. It's obvious that we can't keep these indefinitely, as it also affects database update times which grows with each repository release, repository configs complexity, which includes conditions for older repositories, and ruleset complexity for the same reason, and amount of scroll for project pages. There have been multiple requests to hide EoLed repos from the site by default.
The benefit of keeping these is less clear. My main use cases for it was to
- Show the (still) users of EoLed repositories how much these actually are outdated and vulnerable.
- Keep links to historic build recipes in case these are needed again.
The latter doesn't look valid enough
- It's unlikely that recipes for abandoned ancient software which is removed from the new repos should not be even needed; if some software is resurrected, it will likely need a recipe written from scratch; old recipes would unlikely be compatible too
- It doesn't solve preservation task anyway as packages disappearing from live repositories are just removed
The former still seems valid, but we don't plan to remove EoLed repos right away, and newer, but outdated and vulnerable enough repository releases should be representative enough.
So let's start by removing repos EoLed older than 2 years ago. I'd start specifically with repositories which frequent (thus a lot of) releases such as alpine, fedora and nix, and make exceptions for repositories with extended support or very extended third party support, such as centos. I'd also keep unique packaging ecosystems for now, as these may contain useful quirks and patches.
Latest events such as a migration to larger VPS instance due to diskspace shortage and re-enabling the link checker, which showed that a lot of disk space and link checker capacity is taken by the old EoLed repositories. It's obvious that we can't keep these indefinitely, as it also affects database update times which grows with each repository release, repository configs complexity, which includes conditions for older repositories, and ruleset complexity for the same reason, and amount of scroll for project pages. There have been multiple requests to hide EoLed repos from the site by default.
The benefit of keeping these is less clear. My main use cases for it was to
The latter doesn't look valid enough
The former still seems valid, but we don't plan to remove EoLed repos right away, and newer, but outdated and vulnerable enough repository releases should be representative enough.
So let's start by removing repos EoLed older than 2 years ago. I'd start specifically with repositories which frequent (thus a lot of) releases such as alpine, fedora and nix, and make exceptions for repositories with extended support or very extended third party support, such as centos. I'd also keep unique packaging ecosystems for now, as these may contain useful quirks and patches.