On Mon, 2002-01-28 at 17:42, J.A. Magallon wrote:
> 
> On 20020128 Bryan Paxton wrote:
> >On Mon, 2002-01-28 at 14:00, Christopher Samuel wrote:
> >> -----BEGIN PGP SIGNED MESSAGE-----
> >> 
> >> Hi folks,
> >> 
> >> I'd just like to add my voice to the requests to update the APT package 
> >> listings for Cooker - I started playing with apt-get before Xmas and managed 
> >> to upgrade my 8.1 box to Cooker, but nothing since the 24th.
> >> 
> >
> > I vote second for that, I'm very interested in trying out apt-get for
> >RPM. From what I've heard, it seems to work without a hitch (as much as
> >any software can).
> >
> 
> I don't. Instead of trying to integrate an alien in mdk, why could not
> time be spent in building a curses version of rpmdrake, for example ?
> Make mdk tools better than apt, not just clone apt for cooker.
> 

 It's about choice...
Diversity can be a good thing, it often times pushes on project ahead of
the other, and vice versa.
 But let's have a real life example:
Mandrakesoft produces SNF (Single Network Firewall), I can only assume
that their intentions are to take this further. This being a firewall
oriented target, suppose the admin simply wants to strip the GUI stuff
away, and have it a bare bones system (as any firewall should really
be). If he or she was to choose rpmdrake, well that ruins the plans for
stripping the system. Now, I know what you're going to say, "Well what
about urpmi, the backend? It's a CLI, why not use it?"
And my response to that would be that urpmi has a barage (only in
opposite of apt) of dependencies...

[evil7@sQa evil7]$ rpm -qpR urpmi-3.2-2mdk.i586.rpm
eject
webfetch
perl-DateManip >= 5.40
perl-gettext
rpmlib(VersionedDependencies) <= 3.0.3-1
rpmtools >= 4.0-5mdk
/bin/sh
/bin/sh
rpmlib(PayloadFilesHavePrefix) <= 4.0-1
rpmlib(CompressedFileNames) <= 3.0.4-1
bash
perl-base

[evil7@sQa evil7]$ rpm -qpR apt-0.3.19cnc51-1mdk.i586.rpm
rpm >= 3.0.5
/bin/sh
/bin/sh
rpmlib(PayloadFilesHavePrefix) <= 4.0-1
rpmlib(CompressedFileNames) <= 3.0.4-1
ld-linux.so.2
libapt-pkg.so.3.1
libbz2.so.1
libc.so.6
libm.so.6
libpopt.so.0
librpm-4.0.3.so
librpmdb-4.0.3.so
librpmio-4.0.3.so
libstdc++-libc6.2-2.so.3
libz.so.1
libc.so.6(GLIBC_2.0)
libc.so.6(GLIBC_2.1)
libc.so.6(GLIBC_2.1.3)

You can plainly see the difference, (libapt-pkg.so.3.1 is provided by
apt). Minimalization of software is one of the key elements to security.
So, in that respect, which one would be the better choice?
 Furthermore, I've never set up a firewall or any other security related
system that had perl installed on it, it's just bad news (more options
for an attacker). 

 But take a desktop user, they dont' wanna fool around with the CLI at
all, in general, and in general at first. In that case, urpmi and
rpmdrake would probably make more sense for them.

 Sadly, both sets of software fail to do the impossible, making package
management on Linux easy (yet robust). This is where the development of
RPM comes in, and new packaging systems (hasn't been one yet that is
better though IMHO).

One final note, is mandrakesoft ever going to provide it's own repos of
updates? Bandwidth is expensive yes, but look at the redhat route. I
would certainly pay 5 dollars a month for a service like that (I already
pay for redcarpet). The security and reliability of co-mirrors is
ridiculous. Time to take the next step, don't ya think?
 
 
That's my 50 dollars errrrrr 2 cents : ) 


-- 
Bryan Paxton
Public PGP key: http://www.deadhorse.net/bpaxton.gpg

Reply via email to