I do a lot of self-study, and I use Calibre (ebook library software) to organise my reading materials. I used it because my Linux file browser has problems displaying covers for EPUB files. (And a couple of other) I like that my reading material has lots of ways for me to organise it without creating duplicates (tags mainly) search and browse my books.
Then one evening I found myself wishing I could have all my other relevant 'files' in the same system: audio, video, source code, repos, blogs, podcasts, notes etc... Basically a metadata rich database - except not really a database in my case as I need to pull in 'stuff' from all over the place, and not only on one device.
Files are nice, and I hate anything that takes away my ability to organise stuff on a physical level (which, truth be told Calibre stubbornly refuses to let you set up your folder system, so I cheat by using tags) but at the same time, the problem with the skeuomorphic paradigm of files and folder is that there are so many more options open in the digital world than in the real world when dealing with items.
Imagine being able to sit at your office desk and open the drawer to pull out a piece of paper. When you look at the paper it shows you some text. Rotate the paper to the left and the text disappears and a video starts playing, turn it to the right and you hear a podcast, place it on your desk and 4 more pieces of paper appear next to it - or it calls someone relevant to the info.
All possible with digital info, but not with a real piece of paper - yet.
That is the only thing I see as a real negative about files and folders - it may be limiting what we can really do with our digital world because we're stuck in a file-system metaphor created hundreds, if not thousands, of years ago.
Typically when people talk about that, they mean they want to move from an approach where files have a location to an approach where files are queried based on metadata.
So rather than going to /home/kaslai/images/ to (hopefully) see all my images, I'd instead query my filesystem, perhaps like /owner:kaslai/type:image.
The specific syntax would need some serious thought, especially if we want to make it fairly ergonomic to slot in to existing systems.
There are a large number of approaches that one could take. I'm personally a fan of a hybrid, where you could navigate to a location and then make queries on anything within that location.
IMO one of the coolest features that such a filesystem could have is dealing with music. There's no perfect way to map CD rips to a hierarchical filesystem, but that wouldn't be a problem if I could just open a "directory" like /home/kaslai/music/?artist/?album/ which would automatically make virtual directories based on the artist and album, or I could just use /home/kaslai/music?artist=Smash\ Mouth to get a list of all 182 copies of All Star that I have in my library.
Basically the idea is that the OS identifies and indexes all data on disk providing primitives to access content by query rather than only by a hierarchical path.
This applies to user content but also object code, libraries (providing versioning, compatibility, integrity, whatever criteria.)
Can you ELI5 what this means? Very curious to understand ideas on alternatives to the traditional file-system abstraction.