View Single Post
Stskeeps's Avatar
Posts: 1,671 | Thanked: 11,478 times | Joined on Jun 2008 @ Warsaw, Poland
#252
The "Tablet Debian" project needs to have two sub-projects:
  1. Advanced users (both bootable and chroot)
  2. Plug-and-play / Easy set-up (probably chroot / multi-WM via xomap or Xephyr)
Well, I agree with that point. Mostly, my installer is a project to show that installation of Debian on a tablet can happen without distributing around big installation images.

For my sake, having a UI on top of Installer where you can choose "Easy setup (chroot?)" and "Advanced users", where different targets can be chosen, and things just work, is of course the optimal solution.

[*]installing as little as possible "out of the box"; the user should start with a basic debootstrap rootfs, then they can add "layers", such as "(1) make bootable," and "(2) add xfce4 window manager"[*]backwards compatibility with plug-and-play Debian; users can install plug-and-play and then, when they're ready, "upgrade" their rootfs to the advanced flavour
I agree with this and that's how it is right now in the installer too wrt my "environments". The main itch I had regarding beta3 was that it was very much a single type of environment-image and not very flexible to build "new" environments around.

It would be ideal to be able to start out with a chroot and then click a menu item and it would upgrade it to a bootable Debian, by just running apt-get install nit-boot-support or whatever in the background.

My interest, as you may have noticed, is the plug-and-play flavour.

The highest priority in this is to ensure that it "just works." Someone with no Linux background can buy a tablet at Best Buy, come home and set it up, get the hang of OS2008, then install Debian within the hour.
This one I might be a little curious about, - ofcourse, installation process should be possible to start right after getting the tablet - not sure installation process and downloads and such takes less than an hour though

All technical aspects are servants of that goal. The most important question to ask when working on this project is, "will this make things easier?"
Atleast compared to my work with beta3, I've found that having the different boot tricks and user experience seperated in packages, has made it much easier to refactor the system and alter the installation image at will when needed.

Admittedly, distributing tar.gz's or ext2 images of Debian is "easier", and this would obviously still be possible to do even with installer, but using the installer and doing things in .deb's for this gives the flexibility to instantly cook fresh images when the underlying Debian armel port changes or we need to change something about the UI.

So, we can make "easy" plug-and-play images (downloading big tar.gz's from the web/torrent) that are initially made by the installer and simply wget'ed down, or "flexible but still easy", installer that sets up a partition on a MMC and debootstraps the system up to your choice, and allows you to "upgrade" to Booting Debian easily, or "advanced", where you install a full debootstrap with the intended booting/chroot environment on top, etc.

There should never be a point when the user has to touch the command line. Every choice should have a gui menu.
Are we speaking as part of installation or part of UI, or both?

Also, the impact on the existing OS2008 should be as small as possible.
Don't install anything that isn't absolutely necessary. The project should have as few maemo dependencies as possible (preferably none), and it should be dead-simple to uninstall.
My installer has (pre-startup) dependancies of python, screen and ncurses-base (with an actual UI this might not be needed, but other packages like pygtk would be..)

To install Debian, it requires wget, subversion (getting down packages from svn), binutils (debootstrap), libslang2 (cfdisk), e2fsprogs (mkfs.ext2/ext3), cfdisk, bash3 (debootstrap), and debootstrap.

Excepting cfdisk and bash3 this is retrieved from repository.maemo.org (+ /extras)

Also, since new users have higher expectations, we should focus on speed and optimization of applications.
That's obvious, since we are both when booting and chrooting interested in the user experience.

[*]making sure things are very flexible (I want to boot and chroot to my Debian rootfs... in fact, I want to be able to choose which of my three Debians I want to mount and use!),
Obviously also easy, if we do it with packages and dependencies.
 

The Following 2 Users Say Thank You to Stskeeps For This Useful Post: