Re: Bug#438665: Orphaned packages with quite some users
Bart Samwel wrote:
> tag 438665 wontfix
> merge 438665 445900
> Clint Adams wrote:
>> reopen 438665
>> On Wed, Oct 31, 2007 at 06:22:11PM +0100, Luk Claes wrote:
>>> That sounds like the Dependens should be a Recommends, if so please file
>>> a bug for it.
>> It doesn't look like #438665 was actually fixed.
> OK, then I'll tag it wontfix and merge it with _yet another_ reporting
> of this same issue (which was also tagged wontfix), instead of closing
> it. Discussing this issue (including other versions of it reported about
> every month) has been taking up a significant part of the acpi-support
> maintainer load for quite some time. Raising the issue _yet again_,
> especially in such a high-profile forum such as this, will cause much
> higher-priority issues to be stalled _yet again_. I'm afraid I'll be
> using my spare time yet again to discuss a tiny amount of disk space
> instead of fixing people's laptops that don't work. To reiterate the
> arguments made in the various bug reports: the dependencies are
> currently _not a bug_, they are a _requirement_, hence the Depends. They
> are a necessity for the stated goal of the package: to make all laptops
> "just work". The current state of the surrounding infrastructure simply
> doesn't allow for a different solution. If somebody can come up with a
> bit of infrastructure to fix this up that doesn't have any drawbacks, be
> my guest and submit a patch! :-)
> Sorry if I sound a bit bitter, spend a couple of months in gulag
> acpi-support and you'll understand. ;-) I'll add a couple of actual
> technical notes, for those who are interested:
> 1. In response to this bug report, the dependencies on X clients were
> reduced significantly to include only the package that contains xset.
> This was the part of Depends that was actually superfluous, and it
> reduced the depends load significantly. This was the reason the bug
> report was closed instead of tagged "wontfix".
> 2. What we may actually require is a detection system which triggers the
> installation of packages at installation time, based on hardware
> detection. Something like this was discussed in more detail in #445900.
> In the absence of a dependable non-overkill (not larger than the
> packages involved, not a much higher maintenance burden) system that
> does that, or Recommends which behaves exactly like Depends (like some
> package managers treat Recommends, but not all), I'd prefer to stay with
> plain Depends to actually get laptops which "just work". Unless somebody
> comes up with a patch that guarantees that it doesn't break more than 1%
> of the hardware that we support, of course.
> 3. Regarding the toshset package: I have a working Toshiba Tecra 8200,
> one of the models covered by toshset. I may be biased, but as far as I'm
> concerned the stuff is still useful. :-)
Please do adopt the toshutils and toshset packages so people don't need
to reiterate the same discussion over and over again as you say
yourself. Thanks already for taking care.