En réponse à Leon Timmermans <openbeos@xxxxxxxxx>: [snip] > > > It will probably be hard not to allow anyone to send messages to > other > > > > > > ports, but at least the clone_area() part should be pretty easy to > > > solve ;-) > > > > > > > I'd go for adding a permissions byte to ports, areas and sems... > > just good old Unix semantics, with a umask-like variable > > to set default behaviour. > > (a correctly implemented umask... R5 implements it in libroot > currently, > > so fork doesn't inherit it :^) > > Permission bytes are really outdated. > Most Multi user OS'es are slowly going to ACL's , it really is a more > flexible solution. > permission bits really are too limited! > We'd better skip that phase. > I know I may be "old school" (hmm I'm only 24, but hey...), but I don't find that many pros to ACLs... Anyway I don't mind having ACLs implemented in the filesystem (even, the attributes really make a nice place to put them (and the linux proposed implementation also implements filesystem attributes on purpose). But I don't feel ok adding ACLs to areas, ports and semaphores... it's really overkill and wouldn't just work IMO. On the opposite adding a perm byte and checking perms accordingly to UIDs/GIDs shouldn't impact performance that much. Just my (EPERM) cents. François.