So what do you call a RO snapshot? Or do you now need to write the prefix "RO" and "RW" everywhere when referring to a "snapshot"?
How do CLI commands work? Will you have "btrfs snapshot" and then have to always define whether you want RO or RW on every invocation? This smells like git's bad front-end CLI porcelain all over again (regardless of how nice the back-end plumbing may be).
This is a solved problem with an established nomenclature IMHO: just use the already-existing nouns/CLI-verbs of "snapshot" and "clone".
> […] is clearer because a snapshot generally doesn't mean simple duplication.
A snapshot generally means a static copy of the data, with bachefs (and ZFS and btrfs) being CoW then it implies new copies are not needed unless/until the source is altered.
If you want deduplication use "dedupe" in your CLI.
> There should be no read-only snapshot: it's just a writable snapshot where you don't happen to perform a write
So when malware comes along and goes after the live copy, and happens to find the 'snapshot', but is able to hose that snapshot data as well, the solution is to go to tape?
As opposed to any other file system that implements read-only snapshots, if the live copy is hosed, one can simply clone/revert to the read-only copy. (This is not a hypothetical: I've done this personally.)
(Certainly one should have off-device/site backups, but being able to do a quick revert is great for MTTR.)
