On Tue, Feb 05, 2002 at 07:57:15PM +0100, Josip Rodin wrote: > This Subject: is offtopic... :) > On Tue, Feb 05, 2002 at 12:27:22PM -0600, Manoj Srivastava wrote: > > >> > > The /etc/init.d scripts should be treated as configuration files, > > >> ^^^^^^ > > >> > 10.3.2 doesn't say you must give the administrator some way to > > >> > configure the script. It says the script must be handled like any > > >> > other configuration file -- either it's a conffile, or it's handled > > >> > by the package scripts. Simple as that. > > >> Ah, but it doesn't say "must", it says "should". > > So it is not a RC bug -- but still a bug. > > Hamish> That may be a mistake in policy then because in practice it is a must. > > You mean this should be an RC bug? Why? > This is actually an RC bug, because if one doesn't mark an init.d script as > conffile and doesn't install it manually from postinst, they let dpkg > happily overwrite it on each upgrade and thus break the following rule: > > Configuration file handling must conform to the following behavior: > > * local changes must be preserved during a package upgrade, and > > * configuration files must be preserved when the package is > > removed, and only deleted when the package is purged. At issue here is whether a startup script really counts as a configuration file. If a script is in /usr, we don't expect local changes to be preserved on upgrade. I don't believe that init scripts meet the definition of a configuration file by any stretch of the imagination. Still, we have to face the truth that many users will think they can edit these files simply on the basis that they're located in /etc, a conclusion they arrive at by reading our policy and the FHS. Now, this may be a /flawed/ conclusion, but in the real world, many users aren't going to distinguish between "All configuration files are in /etc" and "All files in /etc/ are configuration files". Which means any maintainer /not/ tagging a file in /etc as a conffile is sooner or later going to have to deal with upset users filing bugs about it. Is it an RC bug if a maintainer tries to take the moral high ground when giving in would take up less of his time, result in fewer bug reports, and cause fewer problems for users? Steve Langasek postmodern programmer
Attachment:
pgpfg4EyC1jUb.pgp
Description: PGP signature