2024 Roadmap / Webi "v2"

[coolaj86]

Webi has become far too complex.

I wanted something with near-0 magic - which is why I opted for shell scripts.

But as I've added one case, and then another, and then another (and pivoted from the helper script to the webi command), it's bloated way too far.

There is a lot of necessary complexity, but there's also a lot of leftovers from the learning process.

My thoughts on fixing that:

• webi needs to be its own command, not embedded in the bootstrap
  (less scary, easier to lint, more fun to add --whatever too)
• webi should be offline-first
  (ex: switch versions, reinstall most recent version)
• and yet... everything that can be decided ahead of time (online), should be
  • webi_download should be templated based on curl and wget
  • webi_extract should be templated based on tar, xz, git etc
  • add new APIs to support newer webi command
• rename bootstrap.sh to curl-pipe-bootstrap.sh for clarity
• rename template.sh to install.tpl.sh for clarity
• link to generate script from website (since the one in the repo may be more abstract)

This shouldn't really be Webi v2.0.0 - because it should be a smooth refactor with additional API on the server-side, not breaking changes, nor a from-scratch rewrite, nor a philosophical shift from the original goal.

Thoughts

• Git UA? https://stackoverflow.com/questions/29117355/git-clone-how-to-fake-user-agent