Hi everyone I tried to gather the opinions expressed previously in this discussion and now I would like to integrate them with data from the Debian and Ubuntu Popularity Contests [1][2] in order to avoid biased proposals.
My first proposal was for the task-kde-desktop and I thought it was useful to understand the criteria that were followed for the inclusion of applications in task meta-packages. Therefore I took a look at the source of tasksel-3.39 to understand how it is organised and for guidelines. A part of a readme file suggests a criterion, it is for new tasks but I think it is useful nonetheless: >Care should be taken when adding new tasks to ensure that the new task is >suitably generic -- it should be something of value to a large number (at >least 25%) of our users; and 90% of our users should be able to understand >what the task is from only its short description. It must not perform the >same general purpose as some other existing task. It must contain packages >that are the ones in most common use, and software that is of the best >perceived quality. I make the following assumption: if a task, like task-kde-desktop, is something of value to at least the 25% of the users, then also the applications included in said task should be something of value to at least the same percentage of users. Excluding of course core applications and small ones that are evidently useful or integral part of said task meta-package. Let's try to make a first tentative with what I proposed in a previous message and see if it make sense: in task-kde-desktop change the dependency of gnome-orca from recommended to suggested. Popcon results: | distro | package | inst | vote | old | recent | no-files | |--------+------------+---------+------+---------+--------+----------| | Debian | gnome-orca | 80219 | 3611 | 61008 | 15536 | 64 | | Ubuntu | gnome-orca | 2212520 | 1166 | 2209850 | 659 | 845 | (see [1] to understand the meaning of the columns) Number of submissions considered in Debian popcon: 197536 Number of submissions considered in Ubuntu popcon: 2766185 Summarising, in Debian: - 40.61% (80219 people out of 197536) installed this package - 9.72% (80219-61008=19211 people out of 197536) installed this package and use it regularly - 1.82% (3611 people out of 197536) use this package regularly in Ubuntu: - 79.99% installed this package - 0.09% installed this package and use it regularly - 0.04% use this package regularly There is a discrepancy between the number of users who installed the package and the number of users who regularly use it, so I wonder if popcon considers as "inst" only the installations requested by users or also installations performed automatically because of dependencies (e.g. task-kde-desktop --> gnome-orca). Given that, I would consider the percentage of people who use the package as a more reliable data. So, because it is evident that the percentage of people who use gnome-orca is under the threshold of the 25%, and because it is not an integral application of the KDE Plasma desktop, this package should not be recommended (installed by apt tools by default). Proposing it as suggested package instead of recommended seems to be reasonable. I took gnome-orca just as an example, it is recommended in every task-*-desktop and maybe there is a reason. However the above evaluation can be extended to other packages. What do you think? Any help in understanding data correctly and drawing conclusions is appreciated. Regards Francesco [1] http://popcon.debian.org/ [2] http://popcon.ubuntu.com/
