This is a request for comment on a basic hardware and software assignment method.
The simplest minimal applied version would be master drive OS and slave drive bearing data.
The next higher levels would include replicated data drives. Not needfully in the same machine or even at the same site.
So how would you set up an install so the "OS" is on say-HDA and the "Data" on HDB" ? With an implied or even explicit addition of HDC thru HDX as either permanent mirrors or detachable external storage backup. Imaging a data drive is less of a challenge than a live OS.
Where this differs from past similar schemes or applications is in the division of OS from Data. The most immediate impact could be for users of multiple OS varieties. The result is a "one user data library - OS Agnostic to the computer accessing it model.The OS needs updating or the drive it is on crashes- your data is safely NOT on the same drive. There are uses for the now open space on the OS drive too. Like process journal buffers as Reiser speculated on. Or pointers to replicant drives holding spatially dispersed copies. EX: /user on a box off site.\
"The Zen of Everything being a file" so far has not covered the WHERE we put the files!
Do note that while Linux has had the "ability" to make all this seem trivial- for whatever reasons we seem to not do so. Comments of a value adding nature are welcomed. Open comment form me to the list: Some projects seem to never get past a " "we could- but why?" stage. I'd like to think KCLUG could do better..
" If your OS drive dies- you might be rather glad your data is on at least two other drives"