How to interpret F18 Blocker criterion

[Adam Williamson]

On Tue, 2012-11-06 at 21:19 +0000, "Jóhann B. Guðmundsson" wrote:
> On 11/06/2012 09:02 PM, Adam Williamson wrote:
> > On Tue, 2012-11-06 at 20:34 +0000, "Jóhann B. Guðmundsson" wrote:
> >> >On 11/06/2012 08:00 PM, Adam Williamson wrote:
> >>> > >Fact is, a_lot_  of people still dual boot with Windows, because they're
> >>> > >not sure they want to switch 100% to Linux, or they still need to run
> >>> > >some apps on Windows, or they want to play games, or whatever. Is anyone
> >>> > >seriously doubting it's a common and important use case?
> >> >
> >> >Arent users on OS-X and Apple hw doing the exact same thing?
> > Sure, but there are far fewer OS X users than Windows users. It's just a
> > numbers game.
> >
> > There is another factor at play there, though, to be fair. When we wrote
> > the criterion, our support for Intel Macs was still pretty much
> > non-existent. That was the time when we explicitly wrote into the
> > criteria that we didn't support Intel Macs. So obviously we weren't
> > going to block on OS X dual boot. That consideration probably doesn't
> > apply any more, though the numbers game still does.
> > -- 
> 
> Well the fundamental question to ask ourselves is if windows or os-x do 
> honor other operating system installed then we should if not we should 
> not regardless of any numbers game you like to play trying to justify 
> this criteria...

Well, I mean, there's no simple answer to that question.

The only time OSes really interact with each other, aside from by user
intervention, is at install time. Neither Windows nor OS X makes any
real attempt to try and 'behave' alongside installs of other operating
systems, if you install them after your install other OSes. That is
known and has been the case for years. It's why the standing advice is
to install Linux after Windows or OS X.

But then, trying to do so is fraught with difficulties. We certainly
don't behave as perfectly as possible in all possible multiboot
scenarios. Multiboot is inherently a hack in the PC BIOS / MBR system,
really. It's all duct tape and gum, and impossible for any OS or utility
to really support all possible cases perfectly.

Once you have two OSes installed and multibooting correctly, it's
usually the case that they'll continue to do so indefinitely unless you
do anything that pokes the MBR, or you start manually mounting
partitions from other OSes and poking them, or you do a major version
upgrade of one and it pokes the MBR in the process of that upgrade. The
OSes themselves will leave each other alone, in general. I don't think
Windows ever, say, just squelches the MBR on update, or anything like
that.

Ultimately the intent of the criterion is that a simple 'do a default
install of Windows, then do a default install of Fedora next to it' case
is fairly commonly seen in the real world, reasonably stable in
behaviour (Windows has not changed in how it behaves in this scenario
over the years), and possible to support. Since it's often done in the
real world, it's something our releases ought to work with. It's pretty
much that simple. The corresponding test case does specify the order of
installing the OSes; we don't support the 'install Fedora then install
Windows' case, and never have (no OS can, really, as Windows will always
just squelch the MBR at install time).
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net