[CentOS] MHTL + legal <- dual-booting <- Re: installing Cents os server 7.0

Fri Jul 3 16:36:00 UTC 2015
Chris Murphy <lists at colorremedies.com>

On Fri, Jul 3, 2015 at 5:59 AM, ken <gebser at mousecar.com> wrote:
> On 07/03/2015 02:51 AM, Kahlil Hodgson wrote:
>> Wow. So many _passionate_ words. Still have no idea what Chris is really
>> going on about.
> Yeah, it's one of those threads with "more heat than light."
> I believe that Chris wants (among many other things) is a CentOS which will
> automatically resize an existing Windows or OSX partition when setting up a
> dual-boot machine.

No, I want both OS's to be bootable after CentOS is installed, rather
than the first OS bootability being broken. Broken is the rule, with
one exception: the first OS is linux without the use of LVM.

And the reason I want it isn't for me, it's for the endless stream of
users who have WTF moments over and over again asking about these
problems, including the OP of this thread. It comes up constantly.
They are so lost they have a difficult time even articulating the
problem with more detail than "it didn't work".

>  I suspect that it doesn't (and won't, and actually
> shouldn't) due to legal and PR reasons.  That is, if some user clicks on a
> button which says "Resize [other] partition", someone somewhere sometime is
> going to complain that a Linux install messed up her Windows or OSX
> partition.  This could lead to a legal and/or PR nightmare for Linux and
> Linux devs.  For these reason I'd think it better that, if you want to
> manipulate, say, a Windows partition, we're all better off if you do that
> with Windows software or at least separately from a no-brain-required Linux
> install program.

That's plausible up to the point where the default installation breaks
the bootability of Windows (and superficially OS X as well which
appears to KP as a result of the installation).

If the idea is to be conservative and protect the user, then the
installer probably shouldn't be stepping on the existing boot loader
without warning.

> Just to respond to one objection in advance (because I'm not going to be
> drawn into this thread anymore than I have), years ago, in the early days of
> dual-boot, Windows put a critical and non-moveable file at the end of their
> partition.  This made blithely shrinking that partition either impossible or
> dangerous-- so quite risky.  And to me and everyone I knew there seemed to
> be no reason at all for this file being where and how it was, except to make
> shrinking the partition difficult.  If you went into Windows, however, and
> knew the three or four steps, you could move this file and then, outside of
> Windows, resize the partition, then proceed with the dual-boot install.
> Now imagine a Linux dual-boot install which automatically resized a Windows
> partition.  Then imagine a Windows software update which included some such
> Gatesenheimer in it.  The above-mentioned nightmare would begin and a lot of
> people would be called in for jury duty over something they didn't and
> wouldn't understand.  Settled out of court, there'd begin a fee on Linux to
> cover damages assessed by the settlement.  The Linux devs would probably
> have to go work for MS at minimum wage to cover or comply with their part of
> that settlement. Absurd and paranoid scenario?  Yes.  Possible?  Still yes.

This is specious because we essentially have that with the current behavior.

The 440 bytes of boot code in the MBR that's the first boot loader
code in the chain to boot Windows is both critical and unmovable. And
yet it's stepped on in a default CentOS 7 installation along side
Windows. And then because ntfsprogs isn't on the install media,
grub2-mkconfig doesn't find Windows, and therefore doesn't create a
Windows menu entry to chainload the rest of the Windows bootloader.

Yes, total corruption of the Windows volume in an fs resize gone wrong
would be worse (unfixable) compared to the likely fixable problem
under discussion, post-install. But I still think for a GUI installer
to get the user into the weeds like this, silently, is a travesty of
UI/UX. GUI programs should not get users into this much trouble. The
installer should do this correctly or it should refuse to install
CentOS or it should warn the user or even advise them on alternatives.
But instead, it does all of the exact wrong things that can be done.

That is not dual boot support.

Chris Murphy