[cumbria_lug] New distro advice

[Michael Saunders]

On Thu, 19 Feb 2004, Schwuk wrote:
>
> Neither can you put Longhorn (nor any version of Windows above 95
> and expect it to perform reasonably) on that machine.

Right, so if Microsoft won't make an OS to run on it, there's no point
in improving Linux so it WILL work on it? Like I said in my last post,
machines from only THREE years ago aren't capable of running a modern
desktop Linux. If you need to spend money upgrading hardware to run
both Windows and Linux, then we've destroyed one incentive to switch.

I'm no luddite and I don't expect GNOME, OpenOffice.org, et al. to run
on a 386 with 4 megs of RAM. But a machine from only three years ago
should be capable enough, and yet we're just chasing Moore's law,
piling on features and paying no attention to efficiency and elegance.

Look at this long term, ten or so years down the road. Right now, we
have the GNOME team (and to a lesser extent KDE) overengineering,
overabstracting, copying Microsoft, whimsically adding code for the
feature or buzzword du jour, making everything bigger, and bigger, and
bigger still. This WILL come back to haunt us. Ten years down the
line, if Linux continues its increasing corporate adoption, people
will have to SUPPORT this code.

More 'moving parts' tends to lead to more trouble and failures, both
in software and mechanical devices. Companies will have to support
this mess of bloated code, and we'll be up there with Microsoft in
terms of bugs and exploits. Wow, it's open source! But all the
developers will be working on GNOME 25.0, very few people will be
working on the old code (how many people are reading/auditing the
source for KDE 1 source?), and 'Linux' will develop a dire reputation.

Look at the problems Microsoft is having with its old and cludgy
codebase. Do we want to be in the same position? You can't just fix
these things overnight -- look at how long it took for Mozilla.org to
clean up the spaghetti code and sort out a usable browser. The more
cruft and redundant features we pile on today, hoping to fix later,
the more we're going to get bitten later on. Nastily.

> Longhorn is years away - do you really believe they are going to be
> supporting todays hardware as anything other than minimum
> requirements?

If they feel particularly threatened by the takeup of Linux, they will
identify our weak points and improve on them (marketing the hell out
of them in the process). On launch, many people were surprised by how
stable Win2k and XP were (well, relatively). They could surprise us
again, with significant performance gains, and we'll be stuck with an
OS that takes three times as long to boot and runs much slower.

Oh, we could get everyone to switch to XFce, but then it doesn't have
integration with Evolution and Nautilus... Ah, might as well start
from scratch then? This is why we have to think about this NOW!

As you rightly say, Microsoft have traditionally been one motor in the
upgrade treadmill, but times are different. Before, they could do as
they pleased, and the masses would upgrade. But now they face
competition -- they'll want an edge over Linux, any edge.

Do you believe that if a modern desktop Linux ran at the speed of,
say, BeOS, it wouldn't accelerate takeup enormously? It'd do wonders.
Home users would be ecstatic. Corporations would be enticed by the
prospect of increased productivity.

> Or home gateway (my home gateway/firewall is running on a box not
> much bigger than that - P200/32MB/2GB running RH8). Try doing that
> with MS.

Oh definitely, on the server/networking front Linux is light years
ahead. But I'm only talking about the modern Linux desktop.

> Until someone takes Linux and does an 'OS X' on it, it's always
> going to be a square peg/round hole scenario using Linux on the
> desktop.

But that's like saying: until someone takes Windows and does an 'OS X'
on it, it's always going to be a square peg/round hole scenario using
Windows on the desktop too.

Gumdrop widgets and snazzy transparency effects do not a productive
working environment make. Many UI researchers have said that turning a
computer desktop into a superficially prettified playground often has
a negative impact on functionality and productivity. A UI should get
out of the way, be invisible, leaving the user to focus on work.

Besides, nobody can do an 'OS X' on Linux. The fundamental
user-friendliness of Apple systems stems from them being single vendor
and single platform; Linux targets eighty-five trazillion platforms
and devices, and as a result it can never be integrated so tightly.
There's no point in imitating OS X's design and interface paradigms as
it's an entirely different market and userbase.

But finally, to return to my original point, unless we start looking
at proper, elegant design and careful coding, we're going to be in BIG
trouble later. Not only could Linux be significantly slower than
Windows, but all the bloat in this code is going to cause security
problems. Entire subsystems and components can't be fixed overnight.

Mike

-- 
Michael Saunders
www.aster.fsnet.co.uk