On Sun, 2006-05-07 at 00:57, Jeff Vian wrote: > > > > Then you have to also adjust grub and /etc/fstab to match on > > every box with opportunities for typos. Normally I just change > > Scripting can handle both the labels and making grub and fstab match > without the typos causing mismatching between the different points. It would, if a tested script came that either worked identically with all distributions or adapted to the quirks of each. Otherwise the script becomes another place you can make the typos. > > > Part of any disk copy plan should take labeling into account in the > planning. If it does not then it is a poor sys admin who gets caught by > surprise. Yes, but the point of a distribution should be to save the sys admin's time, not invent new quirks he has to work around. > Tom made the suggestion that the labels can be changed after the copy is > complete (and before the reboot) which will easily handle the conflicts > by eliminating them before you get bit. Eliminating the labels and using the partition device names that are always unique gets rid of the whole messy concept. > Yes, everyone may get caught by it. So what? Beg you pardon? Can you imagine a commercial product vendor saying that causing problems is to be expected? Yet you probably think that Linux distributions should be usable in most of the same situations. > Everything we overcome is > a learning experience that helps us. Learning to compensate for unique quirks is never a good thing for anyone. It was a waste of everyone's time and will be meaningless when you find a better distribution or this one finds some other quirky way to compensate for barely-deterministic scsi device names. > I got hit by it once. The result? > -- I learned, and have never missed this in my planning since. But you learned something that will soon be useless to know. And how to you know you've found all the quirks yet? Did you run across where the MAC addresses are stuffed so they screw up systems after you move the disks? Maybe there's more that we haven't hit yet. > I think it is time for this long thread about a tiny molehill to > disappear (but that will happen only if the attempts to make it into a > crusade will go away). The crusade is for a philosophy of making a distribution that can actually be used in any scale. I just have a hard time believing that whoever came up with idea of putting labels in fstab actually maintains an assortment of machines. And the same crusade should apply to whoever is deciding to move a random assortment of files to new and equally arbitrarily different locations in each release. Either get all the distributions to agree on a standard and do it at once or leave things alone. And please document everything under /etc/sysconfig - something that hasn't been done since RH9 even though it is vastly different now. Surprises aren't fun for anyone. -- Les Mikesell lesmikesell@xxxxxxxxx