On Thu, Jan 09, 2014 at 10:45:44PM +0100, Michael Schwendt wrote:
On Thu, 9 Jan 2014 22:20:10 +0100, Alec Leamas wrote:
> Yes, still it's an interesting issue... perhaps one count how many which
> actually are installed,
"Installed and used actively" would be more interesting.
Especially with regard to optional plugins, which perhaps are not
loaded/executed at runtime automatically. For example, multimedia users
follow instructions found on the web that lead to installing all codec
packages, whether they need them or not. Watching statistics you might
think "hey, there are WavPack or Musepack users", but maybe they never
use files of that type.
it'd be interesting to know how debian QA takes metrics like these:
http://qa.debian.org/popcon-graph.php?packages=python-bottle%2C+python-fl...
I haven't looked but pretty sure these are not recorded via some
unauthorized callback (being debian and all), perhaps these are just
rough download statistics.
> but many problems also here: users privacy/opt-in,
> easily spoofed, infrastructure.
And it wouldn't force a packager in any way, maybe serve as some minor
influence only.
It would not be the first plugin/subpackage that has been discontinued
during the lifetime of a distribution.
If a package were considered "popular enough", the packager would
not want to upgrade the software to a newer version that removes the
package? There are other more important factors when considering a
version upgrade.
And probably most important, you cannot get an obsolete package to
reinstall automatically once it would become available again. User
would need to take notice and reinstall manually (unless packager
plays tricks or makes it a new requirement).
The package may not come back any time soon, and I actually have no idea
if patching it back from the old sources would be feasible (I haven't
looked to what extent it is broken.) If it does come back in the future
I understand it should be named something else... should that potential
future package _also_ obsolete this one? (I don't think so?)
--
~kad