There is a secondary issue tied to the questions discussed in this thread: is it good strategy to do all package selection quite early during system generation? At present, package selection looks to me as an integral part of the system generation process, "burned" onto the installation media. Must that be so? has this question already been discussed?

Naively, I see an alternative: cut package selection into 2 parts:

(1) have system installation start by configuring a minimum set of packages, those that are strictly necessary to obtain a running system for the hardware and the desktop environment auto-detected / user-selected.

(2) install additional packages that customise the system to the profile desired by the user only at the end of system generation - that could even be delayed to the time the system does its first bootstrap (that could still be done in a way that is transparent to users who want an out-of-the-box installation).

The advantage I see to such an alternative is that it moves the issue of targetting (customisation to the profile a user or an apoplication-specific set of users want) out of system generation and makes the corresponding decisions at release definition easier. To some degree that may correspond to what RHEL - or is it Scientific Linux? - propose.

To play with this idea: the second phase might be done with a tool that could also be available when the system runs under user control, and allow the user to change / enrich his profile (without requiring him to go down to the level of urpmi and installing task-xxx).

A small additional advantage is that this might help to remove some of the "installation-stress" for new users: when they install their first system, the questions on detailed package selection (if the user wants that) would come at a time when the system is basically ready to run and the user has received his first "congratulation" message: when he has stopped being scared.

Please, dont misunderstand: Mageia needs to turn out the first release as rapidly and safely as possible - I do not suggest to modify things now. But - given the present discussion - considering a more flexible way of doing package selection and profiling would open up a different background to the present discussion, it might be helpful at mid-range.

Reply via email to