On Sat, Feb 05, 2011 at 09:11:56PM +0100, Andreas Tille wrote: > Just for the sake of interest: Do you see any other problems than a name > space conflict in the scientific field (which I just assumed for a > moment when I wrote the proposal which is probably not very well > thought).
I'm specifically concerned about confusing users/admins. Let's assume that there exists such a place. There are two ways to put binaries in there: One would be to install all binaries there, the other to install just the ones that cause a name conflict. With the first option, /usr/bin would be populated with symlinks which one first needs to follow, leading to the science bin dir. In there, there is basically no mapping to the package, so one needs to look which binary package puts files in there. This is just confusing. With the /u/l/$package solution, it's directly clear from the symlink in /usr/bin which package the binary belongs to. With the second option, the above problem is not that much of deal, but I'd find it confusing to have some binaries in /usr/bin and some in the science bin dir. It's a kind of "mix". Maybe this is also not that much of a deal. But with my admin hat on I think it is really preferable to have consistent place to look for binaries, that being either /usr/bin or /usr/lib/package, and to keep PATH to default. If every blend would end up putting it's own bin directory somewhere[1] and adding those to PATH, things get messy. But maybe I'm overly concerned in this respect. (I personally think this would violate policy, or at least stretching it to it's limits. One cool thing about Debian is sane defaults, and we should not break with that.) Re-reading my last email, I found the tone more harsh than expected. This was not intentional, I hope noone took offense from that. If this is the case, I'm sorry. And I hope I did better this time. (Not a native speaker, after all…) Best regards, Manuel [1] Like in "Hey, the science folks did that, why shouldn't we do so?!" -- To UNSUBSCRIBE, email to debian-med-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20110205221625.GA3108@woodstock
