Mattia Dongili <[EMAIL PROTECTED]> writes: > On Sun, May 14, 2006 at 12:55:02PM +0200, Hilko Bengen wrote: >> Since the default path is only a minor deviation from upstream and > But it will break setups for people willing to roll their own uml > kernels (or using binaries from sources other than debian's).
I believe that a distribution should make things easy for those users who wish to use the distro's mechanisms, while not making things impossible for users who want to use their locally-built software. The latter group usually tends to know what they are doing; at least I'd expect them to. Therefore, I'd rather burden them to put that extra socket argument into their scripts. >> since it will benefit users who use the mechanisms added by the >> uml-utlities package, I'd rather have the patch applied again. > > why not simply change our script to use the default in /tmp? That one is easy. ;-) Quoting from the FHS: ,----[ /usr/share/doc/debian-policy/fhs/fhs-2.3.txt.gz ] | /var/run : Run-time variable data | | [...] | | Requirements | | [...] | | System programs that maintain transient UNIX-domain sockets must | place them in this directory. `---- Essentially, /etc/init.d/uml-utilities is doing the Right Thing here: uml_switch, started by the init script and running as uml-net has to be considered a system program. uml_switch instances started by individual users, OTOH, are not affected by the FHS. While using /tmp/uml.ctl should work just fine on a single-user system, users on a shared system will want to put the socket and the pidfile into their own home or a subdirectory of /tmp that has been created for just this purpose. But this is the typical "roll their own" case and, again, I'd expect from those users that they know what they are doing. Cheers, -Hilko -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
