On Mon, 2007-03-19 at 10:42 +0100, Alexander Larsson wrote: > On Sat, 2007-03-17 at 22:03 +0200, Baris Cicek wrote: > > I could not find where to post this, so sending over here. > > > > I changed Alex's xdg-user-dirs application to make it using symlinks > > instead of actual directories. Maybe it's something in vain but just > > wanted to put my solution into a prototype so as to be tested. (Maybe > > someone like it). > > > > Actually using symlinks give better handling of localization than using > > --force. It does not touch directories, so can be considered > > theoretically more secure. > > Each application will have to use readlink() to find the actual > directory to use. For sure you don't want have a filename like > ~/.desktop/file.txt to ever be seen by the user, so you must always > store the expanded form. Plus, you need to fallback on a file if > symlinks are not supported. Applications should look at "user-dirs.dirs", which points to symlink. If they won't, they were making big mistake at all, losing necessity of "user-dirs.dirs". Using static non-visible-english-on-disk would make life of packagers easy if dotted directories are standardized. Think about a tarball, created with full of pictures. Packager can use directories relative to ".pictures", and user can see it in his/her "~/Pictures" symlink. How could that possible with translated directories? Fallback is possible and can be implemented. However limiting people using native Linux technologies for those using (no offense) ad-hoc solutions like samba is not a way to go. People share their directories as well, so should we encourage not using symlinks anywhere? I know that's not what you mean. But if someone using Windows servers and using shares with samba, then s/he should bare it's constrains. Using samba as a domain server and file server won't have this problem of not supporting symlinks (samba shares support symlinks). > > So, I don't think this is a better solution. (As I didn't when it was > proposed in the discussions leading to the design of xdg-user-dirs.) I believe that strength of Linux is coming from it's ability to address needs/requests of most possible user base. That's why I wanted to put symlink approach into implementation as well. If we won't use coolness of symlinks for that, where should we use it? (only for library naming conventions?) > > =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= > Alexander Larsson Red Hat, > Inc > alexl@xxxxxxxxxx alla@xxxxxxxxxxxxxx > He's an underprivileged flyboy gentleman spy living undercover at Ringling > Bros. Circus. She's a foxy cigar-chomping hooker on her way to prison for a > murder she didn't commit. They fight crime! > -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
