I want a filesystem that doesn't use c/m/atime. It would use SHA-1 (or other) hash to identify data.
Possible ways to do it:
- hash the entire file (probably not practical unless it could use partial/incremental computations)
- hash every block and then hash the list of hashes
- ...
Programs like Make instead of relying on the mtime would have to keep track of which hashes the sources had for any given target and if the hashes on disk are different then rebuild but at that point no hashes would need to be computed because they would be stored in the file system.