Proposal for solving CD Size problems

Tim Schmidt timschmidt at gmail.com
Wed Sep 27 05:16:22 BST 2006


On 9/26/06, Ethan Osten <senoki at gmail.com> wrote:
> On Tue, 2006-26-09 at 15:42 -0600, Andrew Jorgensen wrote:
> > ther keep fighting to keep it all on one CD than give up and
> > go to DVD.
>
> Let's see what we can agree with:
>
> 1. The size of applications is increasing with time. An average package
> will be gaining new plugins, new features, and new dependencies, in most
> cases without removing any plugins, features, or dependencies to
> compensate.

I don't agree...  Mature applications grow slowly in size, however,
there's a certain amount of application 'churn' where mature apps are
replaced by newer ones.  The newer ones are quite often smaller, and
replace several mature apps at once.  Saving space in fell swoops.

> 2. The variations for computers are increasing, bringing new
> requirements. Five years ago it wouldn't have been necessary to ship
> NetworkManager (although this isn't currently shipped because of
> limitations) or gnome-power-manager; laptops were relatively uncommon.
> Now they (last I checked) have overtaken personal computer sales in the
> US. What happens if UMPCs or something similar become common enough to
> warrant attention?

Eh...  a computer is a computer is a computer.  The computer in my
auto is suprisingly not too different from the computer on my desk
from the point of view of the linux kernel.  Supporting power
management and alternative input methods are things any good OS should
do reguardless of where it's running.

> 3. The number of uses for computers is increasing dramatically. Ten
> years ago it wouldn't have been at all necessary to include a photo
> manager or an instant messaging client by default, simply because people
> didn't use their computers for those things as often. Now we include
> F-Spot and Gaim, not to mention everything else.

Yup.  Computers are usefull buggers.

> 4. The range of hardware is increasing with every year. There was a time
> when you could fit the necessary drivers for target hardware in a few
> megabytes; it's increased to what, now? Notwithstanding standardization,
> it's entirely likely that the number of drivers necessary is only going
> to increase over time.

Ummm...  size of the Linux kernel package in Ubuntu is 21Mb.  It's
maybe doubled in size over the last 7 years I've used Linux.  Nothing
to worry about.

> 5. The expectations for an operating system are increasing. For example:
> regardless of your opinions on the matter of eye-candy, it's only going
> to be more common and expected in the next few years, particularly once
> Vista gains widespread adoption. If Ubuntu can't meet the expectations
> of users, it's going to have trouble gaining users from Vista or OS X,
> or even other Linux distributions without the single-CD scruple.

Not really relevant to CD size.  Compiz is absolutely tiny.

> 6. If Ubuntu is to be a truly international system, it has to include
> localizations for most major languages (and even some of the minor
> ones.) As the number of strings to be translated increases, however, so
> too will the size of these language packs; it may be negligible on an
> individual basis, but when taken with the dozens of languages necessary
> to support and the possibility of more language packs being added, this
> begins to add up.

Agreed.  We _really do_ need to find a nice way to make language packs
super-easily available (they're already easy)...  preferably without a
download.

> 7. The number of things depended upon upstream is only going to
> increase. Things like Avahi are just the beginning; within the next few
> years Gnome is almost certainly going to start depending on something
> like Beagle or Tracker. This is something which is simply _going_ to
> happen; as the needs of users increase, Gnome will have to adapt to
> compete effectively with other desktops. And there's only so long that
> Ubuntu could remove these dependencies before it ends up being
> hopelessly behind.

No big deal.  As mentioned before, apps don't always increase in size.
 Compression gets better, so on and so forth.  It will take work to
keep the base install on one CD, but it's worth it.

> Now, it's clear that at least a few of these things can be solved in the
> short-term by compressing archives, removing other packages, building
> packages without support, or simply ignoring user requirements. But is
> this a viable long-term solution? How long can Ubuntu honestly go on
> 650-700 megabytes alone?

Quite some time I think...  and without building packages without
support / features.

> It's all well and good to say that Ubuntu needs to run on old hardware,
> but there's a certain point where Ubuntu has to look at the situation
> and come to a set of dependencies. It already has a baseline of 256
> megabytes of RAM (which is already a little low; a more realistic number
> is probably 384, having run Ubuntu on 256 in the past) and a CD-ROM
> drive, just to name the most important ones. These aren't dependencies
> which can be maintained forever; eventually, regardless of Ubuntu's
> actions, they'll have to be bumped (imagine running Gnome 2.16 on a
> machine that run Windows 3.11, or even Windows 98!) Is this going to
> happen five years from now, when Ubuntu is far behind all other
> distributions and operating systems in terms of what it supports, or
> will it be when Ubuntu still has the competitive edge?

I have successfully run Ubuntu (Hoary - with Gnome, not Xubuntu or
otherwise) on a Pentium 233 w/ MMX w/ 128Mb of ram.  The user
understood that she could only use the machine for one thing at a
time, and that it would be a bit slow.  She was happy to have a modern
OS with security updates.  I suspect the situation in poorer countries
is quite similar.  'Slow' is relative.

> Make no mistake; it's possible for Ubuntu's main line to bump its
> requirements while retaining support for older hardware. The solution
> isn't to cripple Ubuntu indefinitely, but to create more specialized
> versions. We already have Xubuntu for low-end machines; why not a Gnome-
> or KDE-based single-CD distribution for the hardware in-between? It
> represents maintenance work and possible user confusion, but I don't
> think either of those are insurmountable challenges.

Eh...  better to spend the work compressing the Desktop CD better,
than maintaining X more versions.

> Now, perhaps a single DVD isn't the best of all possible solutions;
> maybe multiple CDs or some other innovation would be better. But
> _something_ needs to be established to deal with increasing space
> requirements. Ubuntu could make up the space increases this time by
> removing some things; what will it be removing next time? At what point
> does removing things become impractical? These are questions that need
> to be addressed, and soon.

I believe the 9% savings from squashfs w/ lzma will buy us another few
releases minimum...  we can start thinking about ditching apps /
moving up to a DVD in 1.5 years...  by which time that may not be
nescessary.

I'm not an insider, and so don't know a lot about the work going on
for OLPC, but assuming it's not too specialized to be usefull for
everyone, it should help out quite a bit in shrinking the footprint of
current Linux distributions...  With OLPC, everyone talking about
userspace sucking, (relatively) recent talk about what API and library
changes to make for Gnome 3.0, and so on, there's quite a bit of work
going on to slim our favorite OS and apps.  It won't be revolutionary,
but a 5% reduction is still quite a lot.

--tim



More information about the ubuntu-devel mailing list