[linux-elitists] Packaging, deps, and office suites
Karsten M. Self
Mon Oct 29 20:21:49 PDT 2007
on Mon, Oct 29, 2007 at 09:54:33PM -0500, Ruben Safir (email@example.com) wrote:
> > > > when their dependant package is removed.
> > >
> > > unless they shouldn't. The whole package management concept is flawed.
> > Ah: "It doesn't work."
> > Rant much?
> > Please see:
> > http://www.chiark.greenend.org.uk/~sgtatham/bugs.html
> Interesting but a flawed paper ;) Frankly, too much is expected from the
> end user.
You're an elitist, right? We expect a lot from elitists.
> Most programs just crash without as much as a whimper. They disapear like
> a champain bubble. Go tell the user to giver you detailed information about
> how that happens. Good luck.
Strace is your friend. If you need network information, run wireshark.
In the case of most Debian install scripts (RH should be similar),
running bash -x on the postinst script and logging the output generally
works. When providing IRC support, to non-elite users, I request a
pastebin of the output. We get a pretty good success rate by this
> > If you can describe one or more specific instances of undesired behavior
> > regarding package management behavior, please do. Better yet, report
> > them as bugs. Effectively.
> I don't want to turn this into a flame war, but packagemanagement is
> just flawed conceptually.
What distro? What specific problem? How long ago? Stable or unstable
> It assumes a bunch of dependencies based on presumptions about the
> distros, naming conventions, and more.
On distributions in which package management _works_, such things are
dictated by policy, e.g.: "3.1 The package name", in Debian, which
includes "3.2 The version of a package".
> Instead of keeping tract of stupid naming conventions and files in a
> database, there already IS a database. That database is called the
> FILE SYSTEM, an organized collection of inodes.
If you're interested in verifying the integrity of something other than
those raw inodes, a packaging system can be most beneficial.
> Programs should LOOK for their dependencies in a variety of rational
> locations. It should know that if lib 1.0 works, then 1.2 should
> work, and be leery or 2.0 only if the programmer says so.
... and it's somehow easier to work directly with a diverse set of
upstream developers rather than put an intermediary set of package
maintainers who understand the current distro's policies, to
<15 lines of sig snipped>
aptitude install signify already
Karsten M. Self <firstname.lastname@example.org> http://linuxmafia.com/~karsten
Ceterum censeo, Caldera delenda est.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 191 bytes
Desc: Digital signature
Url : http://allium.zgp.org/pipermail/linux-elitists/attachments/20071029/39241d20/attachment.pgp
More information about the linux-elitists