Notes:

What if we replaced our directories and file names with a relational database? We could start to understand the relationships between files. “This photo was is an attachment to that email”, “This letter was constructed from that template document”, “This photo is an edited version of that one” We could identify files using the attributes they have.

This simple relational model is the sort of thing I've been pondering. I don't think there is anything radical or new here. For those not familiar with the diagram syntax it roughly means the following:

• A file has an identifier - that's pretty much all. It would correspond with the 'i-node'.
• Files can be related to each other, so we understand that one was originally a copy of another
• A file can be related to a tag. That relationship can carry some additional data
• The type of that data will depend on which tag it is
• Tags can be related to other tags, so you can understand that one is a synonym for another for example

Of course there are all sorts of problems to be thought through. Different people will create tags in their own language or variant of that language. How will files be copied between such systems - would I want the tags you gave to the files when they were on your system?

Security would have to be layered over this system. It is easy to think how this might be done by associating security clearance for a give access to a file with certain tags or groups of tags. However this would require very careful consideration.