Tech Spec, System Installer

[Chris Murphy]

On Feb 22, 2014, at 2:45 PM, Liam <liam.bulkley at gmail.com> wrote:

> Given that the audience for the Workstation is developers, does it make sense to prioritize minimalism?

Yes, because the minimum required should enable the majority to end up with a bootable, working system. I'm not suggesting arbitrarily gutting capability most people really need. But rather focus limited development and testing resources on some common, sane, tested and recommended subset of the many ways of ultimately getting to the same thing: bootable, working system.

> That doesn't, necessarily, mean keep things as they are, but I'm not sure you need to coddle them. It makes sense, to me, to continue offering the two paths but, perhaps, tweaking what they offer/expose a bit.

OK.

> So, assuming someone has Mac/windows machines available, their installers should be used as a guide for user expectations.

The OS X installer is point and shoot. There is no partitioning, LVM, raid, or encryption. A separate Disk Utility application permits some limited resizing, partitioning, raid, and a bcache-like choice. It's used pre-install. Encryption is an option post-install via online conversion.

The Windows installer by default is similar. The custom install path offers basic partitioning. No LVM, raid, or encryption. However most computers come with a manufacturer software restore, which typically doesn't use the Microsoft installer, usually it obliterates everything on the drive in favor of a predefined partition scheme and software set. There are some variations on this, but the most capable and complex I've seen still have fewer testable outcomes than the Anaconda default/easy/Automatic/guided path. By about an order of magnitude.

Fedora probably needs more capability than that. But

> To the point regarding combinatorics and qa, the installer needs to be reliable. If you can't test every scenario (or at least have strong reasons to think untested variants will work) you shouldn't expose them (if at all possible).

Yes or somehow placard or segment or gradually reveal such things.


Chris Murphy