On Tue, 2007-09-25 at 17:19 +0200, Florian Festi wrote:
Jesse Keating wrote:
> On Sun, 23 Sep 2007 17:30:05 +0100
> David Woodhouse <dwmw2(a)infradead.org> wrote:
>
>> That strategy is, quite simply, wrong.
>
> Then work to fix the strategy, don't shoot the tools for following the
> requested script. Dropping snide comments about them doesn't make
> anybody any more eager to listen to you.
The point is that in fact yum is the problem (not the only one). Yum - as an
updating tool - should honor the user's previously made decisions as much as
possible. To be able to do that on a multilib system yum needs to take arch
into account for more or less every decision (especially the arch of the
already installed packages). As yum didn't do that in the past introducing
multilib would have required to rewrite all package selection code within
yum (and some other parts of the tool chain). Instead people came up with
the "install everything" policy with the hope this would hide most problems
of the non multilib aware tools. As we all known this only works for the
simplest cases - not to mention all the other drawbacks that come up on this
list every week (as in this thread).
Umm, you have it backward. When I was originally writing in the multilib
support I asked how it should be done and was told, at the time, by
Jeremy that it should install both of them b/c that's what users would
want/expect. At least, that's what I vaguely recall. This has been about
3 years, now.
So it is not about changing the "default policy" but about having a sane
behavior in our tools that do not depend on any policy but just work [1].
Wrong. It's about the policy.
-sv