> My understanding is that ACLs can be seen as extension over permission bits > whereas the extended attributes can store pretty > much anything. I am not saying which approach is (more) correct though.
Exactly my point as well. Also, it is not correct to say "cp -p now preserves ACL" because it ALWAYS preserved Posix ACLs - it was only not obvious from the man page. Hence my fix there. The only change we introduced recently is that "cp -p" not only preserves Posix ACLs but also NFSv4 ACLs.