Skip to content

Archives

Red Hat 9, and POSIX ACLs

Good techie review of RH9, thanks Padraig. I find this horrifically kludgy, though:

Just a quick observation. The way text editors save files normally, is to create a new file with a temporary random name, and then move/rename the new file to name of the original. Using this technique, if the file being edited has ACLs, the ACLs will be lost. The Vim editor uses libacl to obtain the original ACLs, and then add them back after the save. It is important that other applications that save files in the same fashion are updated to use libacl.

Bad bad bad. Shouldn’t require application code updates like this. I think this is POSIX’ fault. Mind you, according to acl(5), it looks like umask(2) and a concept of parent-directory-affecting-child-nodes’-ACLs seems to apply; so that improves matters a little.

Still, I don’t like the idea of changing something as fundamental as the system calls used to copy and update files in a filesystem, which hasn’t changed in ~15 years on the UNIX platform. I am sure there’ll be nasty side-effects. Maybe that’s why the POSIX 1003.1e ACL standardization effort foundered ;)