Doomsday Engine

What foundation would the backend be built on?

Here's my thinking about the future of package management.

Local package cache. On your disk, a set of .packs is kept locally (in a number of directories). This includes packages downloaded from remote sources, or ones manually put together by the user. At runtime, these will be instantly loadable.

Local metadata database. Information about all the known packages, collected from remote sources and from the packages distributed with Doomsday itself. The implementation of this database could be based on SQLite or just be a simple Doomsday Script record.

Remote repositories. Similar to apt repositories, web servers can set up a specific directory structure that Doomsday can then access to read a compilation of the available package metadata and download individual packages. Interestingly, thanks to (upcoming) FS2 remote folders, clients can access the server's local package cache as if it was a remote repository where packages can be automatically downloaded if necessary.

Virtual repositories. A remote repository that looks like a regular remote repository to Doomsday, but in fact is something different like the idgames archive, where all the WADs are made available as if they were Doomsday packages so that they are accessible via the same mechanism as any other package.