from the perspective of Linux FOSS often playing catchup:
xattr (extended attributes) could be seen as a response to companies building proprietary node-attribute-trees on top of filesystems (databases e.g.), a legacy NIH paradigm in a sense.
Makes me realize that node-attribute trees (databases, HTML e.g.) can now be represented directly on filesystems (xattr, FUSE, tarballs, 7z).
Imagine what a filebrowser or UI's could be...
I think Haiku OS recognized this simplicity paradigm-shift.
Many datastructures could be represented on filesystems, making things like sqlite/katai struct (things I love) redundant:
$ mkdir -p html/body
$ attr -s DOCTYPE -V 5 html
$ cd html/body
$ echo "hello world" > p
$ attr -s style -V "color:red" p
In a (perhaps silly) parallel universe binary data would just mean tar-balls with such open filestructures, manageable by file-browser (extensions).