Are we ever going to address the "This blob took too long to generate" problem, or it going to be stuck in procrastination hell for the next three years?
My suggestion:
We create a new file named RELEASES_EDITION_2021.md, and copy-paste everything from 2021 onwards into that changelog.
Once 2024 arrives, we create a new file named RELEASES_EDITION_2024.md, and so on.
We add a link to these changelogs at the top of RELEASES.md.
That way we have a natural way to split files, and we don't break links to previous releases either.
In short, because releases.rs isn't an official publication of the Rust project. It could be adopted given agreement between the team and the current operator, but that's not the case yet.
That's not completely true. "Deep" links to RELEASES.md#version-release-date would no longer link directly to the relevant heading. Though if the file completely stops ever rendering, the links are effectively broken anyway.
Unless you're suggesting to forever duplicate the contents between RELEASES and RELEASES_EDITION, which causes its own issues due to having to manually update and keep those separate files in sync.
That's not completely true. "Deep" links to RELEASES.md#version-release-date would no longer link directly to the relevant heading. Though if the file completely stops ever rendering, the links are effectively broken anyway.
That might be more clear by looking at the PR, but my point is that the RELEASES.md file becomes a "legacy" file that stops being updated. So existing links are fine, and new updates go into RELEASES_XXXX_EDITION.md files.