Installer Philosophy

I earned my dev wings back in November and joined the installer team. Since then, I’ve written the partitioning code, the dialog frontend (now maintained by codeman), and most of the GTK frontend (Screenshots). While writing the frontends, I’ve come across a question that has yet to be answered: Is the installer intended to make installation easier or faster?

The original intention of the installer was for automated deployments. The way it is designed made it pretty easy to do end-user installs, also. Unfortunately, the end-user aspect of it has overshadowed the original intention. This is partially my fault with the way I’m doing the GTK frontend, but it wasn’t intentional.

I’ve been getting pressure from users/devs to make it more user-friendly and hand-holding. Personally, I don’t think we shouldn’t lower the bar for Gentoo. If a user can’t follow the already very detailed install guide, they won’t be able to admin their Gentoo install. Making the installer hold the user’s hand is only going to flood the MLs, forums, and IRC channels with dumb questions that are already answered in the install guide or related documentation. In my opinion, the user should still have to read the install guide to figure out what to do with the installer.

14 thoughts on “Installer Philosophy”

  1. Agreed. So build into the installer a frame that puts a user through the install guide, or provide a link to each relevant page, as reference.

    It seems to me the installer could go either way (speed/admin vs. easy/user), but the genkernel package, IMO, actually goes farther towards lowering the bar than a GUI installer does.

    The GUI could certainly make it a bit easier to can some (unlimited?) sets of USE flags, CFLAGs, etc. Especially allowing custom sets (i.e. point the installer to a URL early on for a set of “standard” settings used in a multi-user environment.)

  2. Totally agree with John Hoffoss!

    I would not opt for cutting features or speed for the sake of being more user friendly!

    Integrating the manual in some way would be a nice idea though.

  3. Hey! Nice work! I was just wondering, for those of us who like qt, will a qt front end be on the way anytime soon?

  4. Short answer…no. Long answer…if someone wants to write a qt-based frontend, they are more than free too, but nobody that is currently on the installer team is interested in coding it.

  5. Just wondering, when will this installer be ready for the PowerPC arch?, thanks!

  6. The first full release (planned for 2005.1) will be x86 (and amd64) only. After that, I will personally be adding sparc and hppa support. I’ve already got a ppc developer interested in writing the ppc-specific code, also. While it is possible for ppc support to be included in the first full release, it will much more likely be part of 2006.[01].

  7. I’m interested in this project because I am looking for a way to install to a lot of machines. This would make that a little bit easier. πŸ™‚

    Just based on what I have seen from your screenshots it looks like this goes a long way towards “user-friendly”. I agree with you that we still need the users to read the install guide, as that is the key to them knowing how their system works.

    I’ll have to vote for the option of including the install guide. That might be the best way. Hopefully it could even be an online live version rather than including a possibly outdated guide with the installer.

    Just my 2 cents


  8. Andrew,

    Well damn, that thing’s pretty. But, really, I have to agree with your question — what’s the point/goal of the installer project? I think, for most (power)user type people, a scripted install is the most desirable. Like with throck, above, I have the same need at my new job.

    Scripted, easy-to-automate for a mass roll-out type installs would be ubersexy, but wouldn’t _looks_ as ubersexy as your gtk frontend, which really is gorgeous.

  9. Just thought that I would 3rd the vote for speed/automation in the installer. I run 20 Gentoo boxes, and automation is the key. Can’t wait to start testing the projects work!

  10. Hmmm, I know this may be a minority opinion, but I’d like to see some emphasis on ‘easier’ as well as ‘faster.’ Gentoo is a geeky distribution, but it’s also a quality distribution. Making it easier might result in more traffic from “clueless newbie” types, and might make it possible for people unable to maintain a Gentoo system to install it—but it’s also the first step toward making Gentoo more accessible.

    No GNU/Linux distribution, whether intended to be or not, is entirely easy for an end-user to maintain; nobody has really “slam-dunked” the answers to that problem. When better answers come along, they’ll be in the form of applications that Gentoo maintainers can package and Gentoo users can run. I have to think they’ll accompanied by people willing to support them.

    Ease of use in running the installer, accompanyed by the usual warnings that Gentoo is not a newbie distribution, help give those eventual answers a starting point. They’re not going to make Gentoo less attractive to advanced users or create a nasty flood of newbies that can’t be handled relatively easily….

  11. Heard about this project a while ago but never looked into it too much.
    Now I am. πŸ™‚

    While I am fairly comfortable with the current installation methods, I think making it easier and faster is a good thing, or at least can be…
    I am not a computer professional, just a hobbyist, so I don’t have to worry too much about multiple system rollouts but I DO see a big value there for those who do admin networks and so on.
    I think that Gentoo fills a fairly tight niche at this point- intelligent, committed linux users who want speed, control and flexibility. A gui installer won’t change that but it may/should make the “cost of entry” a little less costly (time/effort.)

    More people will be willing/able to give Gentoo a shot, I think.
    Many may stick with it but I expect many who insist on a gui installer will not be happy with Gentoo in the long run…
    Nothing can please all the people all the time, of course. πŸ˜‰

    Bottom line:
    change is a requirement for ANYTHING.
    Some changes work great- we call this growth.
    Some changes don’t work at all- we call this a mistake.
    Sometimes changes simply define who and what we are dealing with a bit better.

    But you gotta try, or else you stagnate and lose.

    Anyway, that is the end of my little spiel; now if I can find an op on the irc channel I will volunteer to help in any way I can (probably just testing as I consider myself a terminal linux nOOb!)

    Keep up the good work and I look forward to seeing the end result.

  12. if the installer basically moved you from item to item, that would be great. you still do it, but it ties things up nicely.

Comments are closed.