[cumbria_lug] New distro advice

[Michael Saunders]

Lo Trevor,

On Wed, 18 Feb 2004, Trevor Pearson wrote:
>
> For a 'home or small office' where funds are tight and
> paid-for-support is out of the question and where the local skill
> set is limited to say one person then FC with 'x' disabled (in run
> level 3 for example) could make sense.

Yes, but I'm not questioning the ability or quality of the distro;
it's the support that's at issue. There is NO GUARANTEE of timely
security updates for Fedora Core, and each release will be EOLed after
5-7 months. This basically means you've got to watch various security
mailing lists constantly, and upgrade the whole distro frequently.

Fedora is an officially unsupported, bleeding-edge hobbyist distro.
Debian or Slackware are far better choices on servers.

> I have installed FC1 on my main workstation. Linux's reputation is
> for amongst other things reliability and security both of which are
> as important if not more important than speed.

That's totally true, but nonetheless it's no reason to neglect one
aspect of an OS. Speed still matters. Imagine if Linux and KDE/GNOME
booted and ran as fast as BeOS... Its desktop marketshare would double
in 12 months, easily, and its corporate uptake would be equally
strong. People love major improvements -- if Lin/K/G/Moz/OO.o
performed much faster than WinXP, there'd be a much greater incentive
for home users to switch. Ultra improved productivity.

> Well maybee Linux doesn't run as fast as Win 95 did but it doesn't
> do all the horrible nasty stuff over and over and over again either
> ? we could fix both problems though !

We _CAN'T_ fix these problems overnight though.

When I first started running Linux full time six years ago (Red Hat
5.1), I assumed that the upstart OS would progress to be the
rock-solid, secure and fast OS of the future. Now we have only two
mainstream distros putting any real effort into stability and QA:
Debian and Slackware (RHEL isn't really mainstream in this sense).

That's horrifying. We must always remember that a newcomer's
perception of Linux problems differs wildly from our own. When we
encounter a bug, we can attribute it to a certain Linux component.
When a newcomer installs Mandrake 9.2, has his CD drive fried, has his
desktop crash, has problems updating packages (which break other
stuff) and has to download over 100 MB of bugfixes, what'll he say?

"This Linux stuff is slow and buggy as hell. Totally overrated."

Worryingly, this is happening RIGHT NOW. Almost all of the big distros
are chucking in the latest, bleeding-edge untested code, and novices
are getting a bad impression. This is a very serious issue.

These are things we can't just fix overnight. More specifically, look
at the KDE project. They've done loads of performance work on KDE 3.2,
and it's noticable, but only _slightly_ better. As time goes on, it'll
get harder and harder to remove the cruft and bloat, and we'll get
stuck in a hole. Meanwhile Microsoft will be tuning, tuning, tuning.

> I have just 'acquired' a p150 with 16Mb of ram and 2Gb of HD space
> (from Mark at SRI, cheers) now I suspect XP is not going to be an
> option, Linux will run on the box and I will be able to use it too.
> Beat that with Longhorn ?

But that's my point. You can't put a Linux distro on that machine that
rivals Longhorn in featureset and capability! At best you could
squeeze on Slackware, but X and even the lightest WM would thrash swap
a lot, and finding appropriate apps would be a nightmare...

Besides, that spec is naturally too low. What got me thinking, though,
was coming across a PC Plus from 2001. In it were reviews of desktop
machines, and they all had 32 or 64 MB of RAM. Now, have you tried
running KDE/GNOME, OpenOffice.org and Mozilla on those? Even AbiWord
and Firebird would struggle.

Just consider that. These machines are only THREE years old, and yet a
decent Linux desktop won't run on it. That's bad. That's terrible.
This was meant to be the liberating OS; the one for everyone.

> Given 512Mb is now circa =A350 and hardware costs are still falling,
> the cost of a min-spec p.c. to run FC is about =A3400 from the likes
> of Novatech compare with the inflated price for a box down at comet
> =A31000ish.

OK, but go out to the third world or a major business with 10,000
boxes running NT4 and tell them they'll have to upgrade them all. Ah
well, they might as well go to XP/Win2k then. That's not how it should
be! Linux desktops would have a MASSIVE market here without the bloat!

> > Yeah, a config system really should eat up 20 megs of RAM...
> >
> Sounds more like MS.

Funny you should say that, because there's a disturbing trend among
the GNOME camp to emulate and imitate Microsoft's practices in every
way possible. It entirely contradicts the UNIX philosophy:

# Elegance -- "Scrap simple and sensible text config files, let's have
a huge, poorly organised and complicated XML database! And an
associated tool that eats up 20 megs of RAM!"

# Efficiency -- "Yeah! Screw that! Did you hear about this feature?
Let's roll it in and see what it does later. There's no problem with
overengineering, feature-creep and bloat. It's not slow on my 4 GHz!"

# Stability -- "Woohoo, GNOME 2.4.0's panel has only five repeatable
crash situations, so let's cut a release! Bugs can fix themselves!"

# Flexibility -- "Nah, we won't let users have wireframe mode
(essential for networked X), nor any way to disable minimise
animation. Sure, people enjoy changing their phone covers, paint
colours, room organisation, and everything else, but WE'LL dictate how
they work now! Mwahahaha!"

This wouldn't be a problem normally, especially considering the
mechanics of open source development, but GNOME is rapidly becoming a
corporate desktop standard thanks to licensing issues and Red Hat.
Looking down the road isn't pleasant, though.

GNOME is already a train wreck of code (even some developers have
admitted as such), and they're piling on anachronistic features with
no idea how they're actually _useful_. So, in ten years time when the
corporate world is standardised on GNOME desktops, we'll have enormous
baggage and complexity to maintain. Welcome to Microsoft Windows.

A lot of people are seeing this happen -- I'm just agreeing. The
GNOMErs are trying to fold in all sorts of stuff from around the OS;
changing a cleanly layered system into one large Windows-esque
monstrosity. This is _unbelievably_ bad in the long run.

> So IMHO on Cost-Benefit analysis it's always going to be Linux slow,
> sloppy overengineered and potentially 45p.

Microsoft have 40+ billion dollars in the bank. They management are
lazy and sometimes stupid, and the company is lead by the marketing
dept., but they employ some of the most talented coders on the planet.

And now they know that Linux is a major threat.

We could go on and on, rolling in inessential cruft, adding features
that seem useful for one day until they're surpassed, not thoroughly
testing and checking code, mixing up the whole system... And Microsoft
launches Longhorn, which:

A) Boots faster than Mandrake/Fedora/SUSE (easy to do)
B) Runs faster (very easy to do)
C) Is more stable than the above (unless QA improves, not hard)
D) Is $49. Ouch.

They have the programming talent and the cash reserves to do this. We
should be focusing on stability, cleanliness, elegance and
performance, because otherwise we could be in for a major, MAJOR
bruise when Longhorn ships. Our mainstream user-friendly distros are
slow, poorly tested and debugged, and often frustrating to newcomers.

/me cracks open another beer

This isn't the way it was meant to be...

:)

Mike

--=20
Michael Saunders
www.aster.fsnet.co.uk