Hi all,
I am examining the workings of a terminal server and have come up to a dead end.
The setup is a central PC with XDMCP getting connections from two peripheral PCs. All PCs have normal Slackware installations (12.0 the central one, 13.1 the others). XDMCP works fine and the users login to KDE. The users have a common directory with appropriate permissions for group work.
The problem is file locking. More than one users can open and work on the same file in the common directory, eg. foto editing the same jpg with GIMP.
Googling around I found that file locking is advisory in Linux, meaning its up to the applications to honour it. Also that there is a mand (mandatory locking) mount option which seems like the solution, but there are warnings [1], [2]. Other finds point to the use of versioning systems (svn, git) to take care of file locking, or network file systems (cifs, nfs): I'd prefer not to burden casual users with having to come to terms with a versioning system, and the network file systems seem wrong since the users work locally.
Considering Linux is a multiuser OS I am taken aback from this apparent lack of file locking support and tend to think I'm missing something obvious.
Has anyone experienced/solved a similar setup? Any words of advice? Many thanks in advance.
A couple of last points. The common directory is on a normal mount (ext3, RAID1), fstab entry:
Code:
/dev/md0 /data ext3 defaults 1 2
And I'm aware of options like TSLP, but I'd rather resolve the issue generically.
[1] man fcntl(2) says "The Linux implementation of mandatory locking is unreliable."
[2]
http://www.mjmwired.net/kernel/Docum...ry-locking.txt