On Mon, 22 Mar 2010, David Kalnischkies wrote: > 2010/3/22 Santiago Vila <sanv...@unex.es>: > > See why I say that "Essential: yes" means "do not remove me easily", not > > "install me if I'm not installed"? > No, i don't see it. > > Your "remove is hard to do, but okay" is absolutely not what essential says:
No, I don't say that "remove is ok", but dpkg has --force-remove-essential. Are you proposing that such option is dropped from dpkg? I see your point about essential status being similar to a dependency, and I see that if apt-get tries to keep dependencies satisfied, it might seem logical to ensure that all essential packages are installed as well. In short: I can agree that apt-get behaviour might be a good "default" behaviour, but it does not follow that this behaviour is the only possible or acceptable one. > I as a package maintainer want to release a package X which depends on > functional on essential package Y. I can DEPEND on the fact that it is > installed, not, maybe, if the user chooses to have it still installed, maybe? Definitely, yes, as a package maintainer you can and should depend on such fact. If the user ever removes an essential package, he deserves whatever bad thing might happen as a result, and this is not your problem as package maintainer, but also it's not necessarily your problem as maintainer of apt. The user has *all* the responsability if he chooses to remove an essential package, because in either case, removing an essential package will trigger all sort of warnings. I am complaining about the fact that apt-get seems to know what he user wants better than the user himself. I hate when packages want to read the user's mind. They usually fail at doing so. I also complain about the fact that you seem to value policy a lot more than the wishes of the user. What the user wants should have higher priority in either case. > [...] > That a system maybe run without this or that essential package only > says that this package is maybe a good candidate to be dropped from > essential - not that it is a good idea to remove it from my system > just because my system seems to work without it. I agree partially. If I remove an essential package and the system runs without it, then maybe it should not be essential. But how I will know how well or bad the system runs if apt-get insist on reinstalling it again and again and again? My complain is that this feature is almost never useful: * Normally, essential and required packages are already installed. * They are never removed "by accident". * If they are ever removed, it's likely that the user decided to remove them. * If apt-get works at all after removing an essential package, then the user can "apt-get install" it himself, he does not need apt-get to do it for him. In this case the feature does not help a lot. * If apt-get does not work anymore after removing an essential package, then the automatic installation of essential packages feature is also useless, as apt-get does not work anymore. So: In which circumstances is this feature really useful to the point that it must be *forced* on the user against his will? (We have already seen than adding packages to the essential package set is easy without it, so that problem is already solved). -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org