[cumbria_lug] Observations

[Michael Saunders]

On Fri, 5 Mar 2004, jenruss+jen at mail.plus.net wrote:
>
> The average home PC has the resources not to care if any given
> program is three times bigger than it needs to be, so if making it
> bigger will make it easier, it's got to be worth considering.

Easier only matters for the first few runs, though. Once a user has
become familiar with a software package, he/she won't need the added
bloat and cruft; equally, it's usually very difficult (or impossible)
to remove the user-friendly components later.

Many current Linux users started off with KDE and GNOME as they became
accustomed to the OS, its workings and its idiosyncrasies. Later,
though, they moved to Fluxbox, IceWM, XFce or whatever to improve
performance vastly -- no matter what you do, you simply can't make KDE
or GNOME run at the same speeds by disabling novice-level parts.

Making a program bigger to make it easier is only a short-term gain;
what matters in the long run is productivity, being able to do as much
work as possible in the shortest amount of time. It's not wise to
bloat an app with 'easy' stuff when it'll be slower to run and use.

Besides, the spec of the 'average home PC' doesn't reflect the real
world at all. Joe Smith with his 3 GHz 1G RAM Dell box may not have
any trouble running KDE/GNOME, Mozilla, OpenOffice.org et al., but a
huge amount of the computers in companies and poorer countries are
unusably sluggish with such apps and desktops.

As I've said before, Linux's adoption rate would improve immensely if
it offered a great upgrade path. If RH, IBM, Sun and co. could go into
a company and say: "Don't spend money on hardware upgrades for XP/2k,
just install our Linux and save!" then we'd be sorted. But the average
box running NT4 or Win98 is nowhere near capable enough to run a
modern desktop Linux, so companies have to buy new hardware anyway.
And if they're splashing out on new boxes, they may as well stick with
Windows for the time being...

Linux would be ENORMOUSLY attractive to businesses if it could just
drop straight in to those NT4 and Win98 systems.

> If Linux is going to spread on the desktop market, it has to become
> more accessible and easier for Average Joe to install and configure.

It also has to offer genuine, concrete benefits to the end-user. When
I first installed Linux in 1998, I soaked up all the improvements
straight away: it was faster, simpler, more secure and more reliable
than Windows. Great stuff. Today, however, modern desktop distros
usually boot slower than XP, have more surface-level bugs and glitches
(Mandrake and Fedora in particular -- see the zillions of updates),
and are becoming monstrously complicated (gconf? Heavens).

This isn't just a babbling M-Saunders Rant(tm) (for once!) -- this is
a real problem and hinderance to Linux's desktop success (both in
homes and in businesses). As a Linux journo I spend an unhealthy
amount of time reading forums, message boards and discussions
concerning newcomers to the OS, and it's getting depressing to read
the growing number of "Linux is really slow", or "It takes ages to
boot" or "The desktop is full of bugs" messages.

And this is why I asked before: what distro can we recommend to
newcomers, giving them a POSITIVE first impression? For all we can
(justifiably) knock the stability of Windows, on the surface XP is
pretty solid and reliable (the underlying foundations are something
else...). Mandrake 9.2, meanwhile, is riddled with glitches, bugs and
problems that make it look like a scrappy, half-arsed toy at times.

Fedora Core 1 is somewhat better in that it had RH's decent QA, but
it's unbelievably slow. Microsoft did some effective work in reducing
XP's boot time -- even though it's still booting once the desktop is
up -- and the current crop of distros are making it hard to advocate
Linux as the "fast" and "clean" OS.

> what's to stop people releasing two versions of their package - one
> with full install assistance, and one with only basic install
> assistance?

Support. Maintaining separate codebases and/or packages is a
nightmare; just ask RH about their policy with GCC "2.96" or SELinux
in FC2 kernels. It's a good idea, but very difficult to implement.

Mike

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