Subject: Unsafe for production use. was Re: etc-update Noob mistake
Innocent bystander @...co.uk writes:
> On Thu, 30 Sep 2004 08:47:21 -0500, Jeff Smelser wrote:
> > FYI, you can change etc-update to confirm if you want it rm/mv the
> > file just like that.. ;)
> Yeah, but like you said, this isn't really an etc-update problem...
No, from reading this thread (linked from weekly news) I'd have to say it's an _attitude_ problem. A n00b writes in with a question about why the software update proceedure he was told to run made his system not boot. A few helpful voices urge the noob not to reformat and reinstall and offer to help in IRC. If only it had ended there.
Cries of "Oh, not that again" and "That one got me once, but you'll learn" and "You've got backups, right?" and even "Just pop the LiveCD back in, and fix it, man" are heard. Explanations of why this is a design fsckup in the OS in question are shouted down by egotistical power-users who want everyone to repeat the mistakes they had to to make, and by people pointing out that the n00b made the wrong choice, tough. Any discussion of how to fix this obviously recurring problem dissappates.
Put me down for fixing base layout, and re-examining how portage (doesn't) handle config files. To see how it could be done less poorly look at Debian policy or FreeBSD's mergemaster Mergemaster does what some posters in this thread seem to think etc-update can do. For instance rather than drawing an integer number line from -1 to -5 and asking the user to guess, mergemaster actually has online help (hit ? [enter]) which explains how to merge or ignore changes.
Also, don't worry about me running Gentoo for production. If I want my weekly software patch run on a production webserver colocated offsite to result in a non-booting system and a trip to the NOC (to use that LiveCD ;) I'll just use RedHat or Windows, thanks. In all my fat-fingered absente mindedness I have never been able to break a Debian or FreeBSD box that badly just by patching it. Wow.
/dev/ROOT / XFS