I think most of my editorial comments have already been raised by Lee.
Some things I think the current proposal does not cover that perhaps it
should:
* The proposal does not seem to cover filesystem choice at any point.
* The network interface manipulation external dependency (DEP-3) does
not explicitly mention DHCP as an option. Nor does it mention
configuration of DNS settings.
* For scripted installs, it is not clear whether it is required to have
a display. I would assume they do not, and the minimum operating
environment specifications seem to reflect this by having separate
requirements for the Runner and the GUI.
* The futural OOBE functionality has overlap with the proposed feature
set (Ch. 2 "System Features"). Will this future functionality just
delay the relevant configuration for later, or does this mean that the
relevant configuration will not be included in the initial Horizon
release?
* FDE is mentioned as a need for "the beginner" user class but is not
brought up again at any other point in the proposal.
* I think Lee mentioned this already, but it would be nice to have more
citations for the shortcomings in the prior art.
* Does any of the prior art allow the user to minimally install in a
dualboot+ configuration? I think this is extremely hard to get right,
and places liability on the installer for possible data corruption
rather than the user, which is why I figure most of the prior art
doesn't do this as far as I'm aware. Maybe you can get around this if
you put a huge "BACKUP YOUR DATA FIRST" screen. Especially with the
large variety of different boot firmwares and boot managers this seems
really difficult to pull off (though in a later section the scope
appears to be limited to only GRUB and syslinux).
* I don't know much about NOOBS but from what I gather it's some sort of
installer - is it relevant to discuss this as prior art as well?
Finally, I think this is the most important point I want to raise:
* I think the proposal should address the relative merits of scripting
installs versus spinning custom images for large deployments. It's
something I thought of several times while thinking about the project.
Max