Re: policy violation and bug reports. - some resolution?
Dale Scheetz wrote:
> It has always been my understanding that a conffile was a special class
> file with respect to dpkg and nothing more. It has special rules for
> replacement, but every time I asked Ian J. I was told that a conffile was
> not the same thing as a config file. If the policy is intended to dictate
> that all conffiles are config files then I am willing to say that the
> policy is broken.
Yay! This seems to completly agree with what I originally said, and I'm glad
you agree and that Ian agrees about the definition of a conffile vs. a
configuration file. [1]
I do think this discussion has brought to light some points that we could
work on. Manoj is really right that it's better to simply touch zero-btye
game score files in the postinst. And we need to look more closely at dosemu
and lambdacore and maybe come up with some way to upgrade parts of them, not
the whole thing as a single conffile (although this seems extremely
difficult to implement).
> This argument has been one of interpretation. This is always going to be a
> problem with the policy. If we choose to live and die based on the words
> in the policy manual rather than making a reasoned decisions, when
> necessary, that potentially violate policy, the distribution will suffer.
Complete agreement.
Also, I think we really need a document explaining the reason behind policy,
to provide some historical background.
--
see shy jo
[1] And I think this difference needs to be highlighted in the policy
manual, which uses the terms interchangebly and vaguely in places.
Reply to: