On Wed, Jan 30, 2002 at 06:52:07AM +0800, csj wrote: | On 29 Jan 2002 11:03:07 -0800 | Dave Carrigan <[EMAIL PROTECTED]> wrote: | | > Walter Tautz <[EMAIL PROTECTED]> writes: | > | [...] | > | > > Just curious to hear other people's opinions on this matter, i.e. | > > don't use devfs. It seems to me the debian kernel should have | > > CONFIG_DEVFS_FS=n. | > | > Some people want devfs. Devfs can't be created as a module. Hence, the | > logical choice is to build the kernel with devfs support. Nothing's | > forcing you to use devfs, even if your kernel has devfs support, and the | > overhead is not very much. | | But, let's say I just want to play with devfs, could I still go back to | my old disk-based setup? Could I still boot my old devfsd-disabled | kernel after using a devfsd-enabled kernel?
The only hiccup I think you'll see if you do this is : Suppose you decide you like the devfs name so you edit your fstab to use devfs names for all partitions. Now you boot your static-dev kernel, and, oops! those device names don't exist. You could, of course, create static directories in /dev and stick inodes there so the old-style kernel works with new-style paths. A better solution is to use devfsd with your devfs kernel to provide compatibility links and keep the old-style paths in your application config files. (this reminds me, one of these days I gotta boot with a floppy and clean out /dev and /tmp since I'm using devfs and tmpfs now) -D -- Micros~1 : For when quality, reliability and security just aren't that important!
