[rudasn]

I'm on and off spending some time on a (probably) completely useless and really bad idea: a filesystem based document "database" with very basic indexing capabilities, in nodejs.

It sounds so horrible, I can't get my mind off it! :P

[scottlilly]

I've been working on something (probably) similar in C#.

The idea is to tag files/documents with attributes like "auto", "insurance", "2022", encrypt the file locally (using your manually entered key), copy it to a network share/cloud storage/etc., and add the tags and document information to a database.

Later on, you can search for documents based on the tags and get a decrypted copy of the file.

[jacknews]

This doesn't sound horrible at all, in fact tree filesystems look like an anachronism from when disk space was limited, and both the logical layer of naming and metadata of objects, and the management of the data on storage, were conflated into a single data structure.

A DB to contain all the metadata, and blobs or object storage for the content itself (perhaps with backup metadata). The database manages the namespace, object storage manages the storage layer.

You can easily and quickly search file metadata etc, there could be multiple copies of the content, perhaps distributed, etc, etc.

[rudasn]

Not similar to what I posted above, but similar to another idea I had but didn't pursue as much:

remote storage with local indexes, where the remote storage only holds encrypted data (can be postgres or s3 or whatever), and the only way to make sense of the data is if you have access to the local encryption keys and indexes.

The local indexes would function like your notion of tags, and also be somewhat encrypted (hashed probably) so you can find your data locally but only pull it from the remote when needed.