Re: 64-bit transition deadline (Re: Etch in the hands of the Stable Release Managers)
On Sun, Apr 08, 2007 at 11:15:56PM +0200, Marc 'HE' Brockschmidt wrote:
> Robert Millan <rmh@aybabtu.com> writes:
> > Does that mean you don't believe there will be such battle, or that you
> > don't believe the predicted date?
>
> I don't believe that there will be such a battle. There is no reason to
> switch from i386 to amd64, no real gain for users.
It isn't users who drive the switch, it's application developers. Eventualy,
the need for a single process to address more than 4 GiB of memory will become
unavoidable. At that point, there are only two choices:
- Use win32 + PAE. This allows your program to run in the vast majority
of computers, and provides the biggest profit in the short term. I hope
this approach will be the most common, and expect it'll sustained for long.
- Use a 64-bit platform. Which one?
- win64 is so utterly broken that none of the former win32 users wanted
to migrate. As a consequence, there's practicaly no userbase.
- x86_64-linux-gnu is a complete product. It has drivers for everything
and a consistent system that works out of the box without 32-bit hacks.
Its userbase (much like i386-linux-gnu) is marginal but still bigger than
win64's.
Those who go the PAE way are totaly irrelevant. Someone (including microsoft)
is going to make a lot of profit exploiting the decadent 32-bit + pae market,
but sooner or later it'll collapse.
Those who go the "clean" 64-bit way will have to make a choice to determine
which will be the dominant OS on this new platform. Their main concern will
be (as always has been) userbase. If our userbase is bigger than win64's
(and that's not too hard), x86_64-linux-gnu will be stablished as the standard
system and the gradual replacement of 32-bit hardware will render microsoft
obsolete.
> I disagree here, because I believe the essay to be wrong and thus
> worthless for further considerations. The basic assumption is that you
> need to move to a new processor architecture because of a need for more
> memory than your current architecture can manage. There are two reasons
> why this argument is simply bullshit:
> (i) There is no real need for more than 2 GB of RAM on a desktop. Even
> servers are usually not in need of such a mass of memory, and this
> will not change in the next 18 months. [1] [2]
8 GiB is not too uncommon for servers nowadays. As for the desktop, it
basicaly boils down to:
- Moore's law predicts when total amount of system memory will surpass
4 GiB.
- Software (thinking of scientific/engineering applications and games here)
tends to accomodate to exploit all the capabilities of your hardware.
- Therefore, Moore's law determines when applications that need more than
4 GiB will appear.
- As long as win64 is so full of crap, most of them will go with pae.
Only a few will go for 64-bit and these are the ones who are going to make
a choice for a new OS.
> (ii) Apple has shown that transparent emulation between two processor
> architectures *can* be an option, and that multi-architecture
> binaries are also possible to smooth a transition. I doubt that
> Microsoft will be able to force all third-party vendors to
> transition to a new architecture at the same time, so Windows
> will be basically bi-arch at some point
win64 is bi-arch already. It just happens to be completely broken as a
platform, because when switching to 64-bit kernel, they lost all the drivers
(too bad, no source code! haha :-)). Nobody wants to migrate to a platform
in which all your peripherials stop working. Heck, in terms of hardware
support it's like migrating to GNU/Linux was 8 years ago!
Also, half of their system (no, not 3rd party apps!) is 32-bit (again, because
of dependance on sourceless stuff. codecs force WMP to be 32-bit, and browser
plugins force IE).
Microsoft themselves acknowledge that their win64 platform is broken by
not providing it in the retail boxes for either XP or Vista.
> and users won't be forced
> to do a hard transition anyway, so there won't be a phase where
> Linux could gather all of those poor, aimless users under its
> cuddly penguin wings.
Sure, no such phase. I wasn't suggesting that.
> Anyway, debian-release@l.d.o is not the place to discuss such things,
> they are not relevant for our decisions about a release schedule, there
> is no actual goal that should be kept in mind for the release planning
> and all of us can simply go on with out lifes.
>
> Please do not answer on this list.
Please excuse me for not obbeying this request, but it seems that it was done
in the assumption that my reasoning is wrong. Since we haven't finished this
discussion, it's quite impractical to resolve it this way since we'd just
keep disagreeing and wouldn't reach any conclussion (if I'm wrong about
all this, I'd like my mistake to be exposed so that I don't have to waste
my time bothering about 2008's doomsday ;-)).
--
Robert Millan
My spam trap is honeypot@aybabtu.com. Note: this address is only intended
for spam harvesters. Writing to it will get you added to my black list.
Reply to: