Yes. One with a generic schema, and to which the user has full access rights.

Fundamentally, it's not the files that make filesystem great. You could devise different models, perhaps a relational one, and paper complexity over with well-designed UIs. They'd probably still be more complex than the filesystem - files are about the simplest data storage abstraction you can invent[0] - but they'd be serviceable, and users would learn.

What makes the filesystem great is that it's an old abstraction, designed in the ancient days back when computing was still about enabling users. Bicycles for the mind and all that. People cared about making things useful to users, instead of just shamelessly exploiting them. So, designed back then, the filesystem grants users the vocabulary to manage their data and freedom to do so, and it's so ingrained that - despite their best efforts - companies weren't able to completely take it away.

Filesystem persists for the same reason e-mail persists. Despite its warts, it's one of those technologies made before the computing industry became exploitative.

--

[0] - Despite the frequent claims to the contrary, coming from the web and mobile world. But guess what, data magically held in app and "shared" by magic isn't easier to understand, it's just not understood at all - users have no mental model for this. The mobile app approach works only because it removed all data management features except the share button.

You are missing the point here by seeing only the technical aspect. You are misleading technical description with abstraction.

A database is hardly if not accessible to the user. The files are an abstraction that enables the user to owns its data. Once you have a file, you can do more than opening it in your app. You can store it wherever you want, edit it with any software you wish, arrange it the way you are confortable with.

As a user, you can’t do that with your SaaS database, you must rely on the « export / share » function of your SaaS provider, hoping it will export all data, in a readable format and that the import function exists and is reliable. You don’t own anything and as soon as you stop paying your subscription, you are stuck with nothing.

If the file roughly maps to a table, the filesystem and filemanager would be the query? To get access to the data, you must first find it.

A filesystem would be like an index on the table.

File formats like DOCX or PSD (fitting extension) are almost impossible to parse 100% correct and render without the (hired) software used to create them. While you may be able to copy your files, without the software they are quite useless.

The file system won't save you there.

You are right and it's a real issue, but you can share them with whoever you want, however you want. You can just keep them and be pretty confident that they will be readable again in years : even if it could become difficult as the time passes, if it is nothing too exotic, chances are you will be able to at least read them.

And I prefer a "95%" correct DOCX or PSD that I can recover and rework if needed than a "0% this App is not available in the WhateverStore anymore".

For popular formats that would be possible, but there are a lot of binary only formats that are completely impossible to parse these days. You're free to copy it but the bits are essentially useless.

Try to load a current ML model 20 years from now. Probably tied to proprietary software and if you're unlucky also hardware (like CUDA)

In fact that is how it goes on IBM i (nee AS/400).

Not only that, but a bad database.