Alternatives to Fedora Core, was: [RULE] APT, YUM...

[Da Worm]

On 7/29/2004 at 6:13 AM M. Fioretti wrote:

> 1.b (Gentoo specific): how do you upgrade gentoo? Can you use and
>    upgrade Gentoo for one year without compiling anything?

Use, yes, upgrade, no.  Portage is great for keeping current, but...

You have to have space to store the Portage tree, and you have to have
CPU to compile new packages.  It does keep recompiling to a minimum,
where it can, in that it only recompiles what is needed to be recompiled.

There are many situations where Gentoo would be ideal, that
do overlap the Rule goals.  However, the single older PC with slow net 
connection isn't really one of them.  For systems like the VumBox project
based on Rule, it would make a great system, because in that environment
you can probably count on one more powerful central server to host via NFS 
(or other network filesystem) the Portage tree.  Plus, you can also use this
server to build the binaries for each workstation, if the workstations do not
even have room for the intermediate binaries.

The biggest advantage of Gentoo is that you can optimize for your system,
and can include only the components you want.  The biggest disadvantage
is the space it requires to hold the sources and final binaries (basically, the
Portage tree and the apps you install).  The Portage tree is the hardest to 
deal with.  A 2 Gig partition (huge, in many circumstances) is plenty to 
build a CLI system (actually, 512 Meg is plenty), but will run out of space 
building KDE, at the minimum level KDE can be built.  Not sure about 
smaller window managers, as KDE is the only one I've tried.  KDE fits fine 
in much less than 2 Gig, but not KDE and full sources required to build it.
Another problem is time.  I built a Gentoo system on a Sparcstation 20 with
a 60 MHz processor and 4 Gig drive.  The CLI portion took about a day,
KDE took 10 days, non-stop.  Not many people are going to want to wait
that long.  I did it to see if it could be done.  Well, yeah, it can, but why?

One possibility is to use Gentoo to build several "mini-distros".  The problem
is that it would require several of them to cover a large percentage of potential
users.  The kernel isn't that big of a deal, there you have loadable module 
support, and can pick and choose among precompiled modules.  But beyond 
that, you still have all the packaging and dependency problems.  Again, here 
the guys who do VumBox type installs have an advantage, as they can 
dictate what gets installed, but that won't satisfy the average potential user.

Someone commented about compiler options in Gentoo.  That could
be a problem, but isn't really.  Gentoo's handbook actually warns you
agains too many "optimizations" as that can break systems in ways that
are almost impossible to debug.  The goal isn't to tweak the optimizations
so much as to use the best optimizations for your system.  Generally, this
means -O2 as the default optimization level for gcc, and setting your march
variable to match your processor, and perhaps -omitframebuffer and -pipes.
(Those from memory, probably not quite right.)  The power and space savings 
come from what support is included or excluded, not from optimizations. 
For instance, not including Gnome, or not including ESS when you have
ALSA compatible hardware (prepackaged distros will usually compile in
every possibly layer since they don't know what you'll be installing to).  With
Gentoo, you MUST know what you have in order to optimize, from the 
kernel all the way up.  You CAN install pure binary (use premade stage 1
through 3) but you get little optimization over other distros.  To truly
optimize, though, you have to start with stage 1 and compile the rest,
knowing up front what you have and what you need.  This is certainly
beyond someone just wanting to try it out.  I've only done about 10 installs
with it, and I am still somewhat clueless about certain things.  I'm
learning with each install, and that's what keeps me going, but that is
my motivation, not just trying to get a usable system up and running.

I wish I had answers that would satisfy all potential users, but the best I can
come up with requires someone, somewhere, to have at least one "big
box" to do the grunt work for each and every install.  Practical in some 
circles, but not in most, I'm afraid.  The end result is wonderful, but too 
many people can't even start.

Jeff.



_______________________________________________
Original home page of the RULE project: www.rule-project.org
Rule-list at rule-project.org
http://rita.choice-secure.com/mailman/listinfo/rule-list_rule-project.org