Tech Spec, System Installer

[Chris Murphy]

On Feb 20, 2014, at 2:28 PM, Adam Williamson <awilliam at redhat.com> wrote:
> 
> You might need to provide more details and background in posts like
> this, Chris, considering the context. I don't think the desktop team is
> as familiar with the ins and outs of installer partitioning as the
> anaconda and QA teams. They don't deal with it every day. :)

OK I'm not sure how or what to provide that wouldn't also be obscenely verbose.

My premise is that the present installer paths (Automatic and Manual) do not constitute limiting the user interaction to the minimum. But maybe all or most WG members consider the installer already meets these requirements?

For comparison, by at least two orders of magnitude, the Windows and OS X installers are more minimalist. They each offer a handful of installed outcomes (including dual boot), whereas Anaconda Automatic/guided path alone offers dozens of outcomes, and the Manual/custom paths offers hundreds possibly infinite.

QA presently lacks the resources to test all possible installer outcomes. In fact it's likely that most outcomes aren't tested, and even if not certainly a significant amount aren't tested. This directly impacts user experience because for those untested outcomes they are actually the tester, possibly for the first time. That's also not minimum contact especially if they get a crash and have to start over, or the resulting installation doesn't work. I think it's sane for users to expect everything presented in a GUI installer is at least somehow minimally tested, yet QA simply can't make that claim right now. 

So should the installer be permitted to enable users to created untested (actually untestable in some cases) outcomes? I'd say no, but that's my own bias. So what's the WG's bias? What do they mean by "the minimum" for user interaction with the installer?


Chris Murphy