It gets annoying though when you have just single invoices or documents that don't quite fit in your structure. I eventually started throwing all received documents in one folder and used the name as tags basically. E.g. 20210214_plumber_invoice.pdf or 20181204_someshop_invoice_playstation.pdf... (And yes I use the filename for tagging because I don't trust e.g. Windows tagging system to be there forever or be copied properly onto other operating systems.)

Please can I have all the Plumber correspondence from Joe the Plumber, regardless of year or document type?

Do you mean Joe the Plumber from London, UK or Joe the Plumber from London, Ohio ?

We all know how this ends up. It ends up being like Google where the search engine uses word embeddings and the like and removes word from your search queries or replaces November by December because they are both months so you can substitute one for the other right ?

Don't care, just give me both. Since I've never been to London Ohio I don't think it will be a problem.

Modern shells have double-star globs for traversing arbitrary numbers of layers, so I would do something like this with:

  ls **/*Plumber*

Thanks, you just got me all the docs from Samantha the Plumber and Amit the Plumber too!

And tags don't help with that, when you just tagged everything with "Plumber".

They could help if the tags were 'Occupation = Plumber' and 'Name = Joe'. Now a search for all files where both tags are present will get you Joe the Plumber's invoices. If you want everything from any plumber or from any person named Joe, then just leave off one of the tags from your search. It is very much like when querying rows in a relational database, just adjust your WHERE clause.

You just summarised the whole argument : you can go very far with a well structured relational database.

I agree. Unfortunately, right now the 'well structured relational database' is completely separate from the file system. Didgets was designed to combine the two into a single coherent system so that you can't update one without the other. By 'combining' doesn't mean I did what WinFS tried to do and just take a filesystem and a database and stick them together somehow. I built a completely new system from the ground up that incorporates traditional filesystem features (block allocation, stream management, metadata control, folder hierarchies, etc.) with solid relational database features (schema, tags).