On Tue, Apr 9, 2013 at 9:01 AM, Cristian Morales Vega email@example.com wrote:
On 9 April 2013 13:58, Cristian Morales Vega firstname.lastname@example.org wrote:
On 9 April 2013 13:47, Greg Freemyer email@example.com wrote:
How is the decision made to have a recommends statement made?
== my situation
I packaged fio for 12.3.
Fio is a performance testing tool for disk subsystems. One of its features is it can produce plots viewable via gnuplot so there is a recommends gnuplot in the specfile.
I've just had a user raise the concern that on a server gnuplot brings in a ton of desktop related packages. Further he argues he would run fio on the server, but gnuplot on his desktop.
I can see both sides of this, but I am inclined to remove the recommends statement.
Why? The whole point of being a *recommendation* is that the concerned user can decide to ignore it if he wants to.
Or "suggest" it instead of recommending it if you want to change the default action.
Okay I changed it to Suggests in my home package.
I have to say, that is almost like not having it in the specfile at all from what I see.
zypper seems to totally ignore "suggested" packages and I don't see a way to make it pop them up to the user interactively.
Yast2 also seems not to have a way to install suggested packages.
I did see that in Yast2 I can look at the dependencies tab for fio and see that gnuplot is suggested, but how many users are going to do that.
If opensuse is going to have suggests be a legit keyword in the specfile, then zypper and yast should have some way to present that suggestion to the user.