The Filesystem

The file is just an array of bytes (did I say this?)
Metadata is unrelated to the file's contents and lives elsewhere
The fs includes an "inode" struct, for metadata (size, owner, on-disk placement)
Inodes are enumerated: device+inum identifies a unique file
For names, we have directories: just lists of names and inums
Names are completely arbitrary strings of bytes
Not really: the separator and null-byte are forbidden
This overly-simple approach brings in interesting side effects
You can rename or remove a file that is in use, and it works
Files can have several names, and it works
Files can have no name at all, and it works
The filesystem ignores local-language encoding, and it works
Names are case-sensitive (and anydifference-sensitive)

So, the filesystem is just a database, whose data items are files
And system programs are the queries acting on it