Having the metadata.json on the server would allow the install script to download the metadata and then figure out which tarball to download from there.
That would save us having to run a sandworm instance that does the same thing but in OCaml and make it easier to scale and test as all the data is static. There is no Docker instance to start and restart and rebuild which should reduce the number of moving parts and things to worry about.
That way the project would shrink to just building the static assets and uploading them, from where a static web server can serve them (and they can be cached by a CDN).
Having the
metadata.jsonon the server would allow the install script to download the metadata and then figure out which tarball to download from there.That would save us having to run a
sandworminstance that does the same thing but in OCaml and make it easier to scale and test as all the data is static. There is no Docker instance to start and restart and rebuild which should reduce the number of moving parts and things to worry about.That way the project would shrink to just building the static assets and uploading them, from where a static web server can serve them (and they can be cached by a CDN).