a bit policy, a bit devel [was: Re: Find the display]
Manoj Srivastava <srivasta@debian.org> writes:
> >>"goswin" == goswin brederlow <goswin.brederlow@student.uni-tuebingen.de> writes:
>
> goswin> Yes, pretty strong words. Hope they wake up some people.
>
> I suggest you give it a little more thought, and a little more
> analysis, if you seriously want to do this.
>
> goswin> It was a lot of work to get to debconf (also not quite the best
> goswin> solution in my eyes as a perl hater)
>
> A perl hater?
Yeah, I hate perl. Its easy to create bugs and hard to fix them.
>...
> goswin> I also know that policy says otherwise (as quoted in another mail),
> goswin> but I think it is time to change that for woody. User interaction
> goswin> should be discouraged by policy, esspecially for such stupid reasons
> goswin> as bad code.
>
> Please substantiatre this statement. Or are you just
> considering the sawfish fiasco here?
The bad code part is ment for such thinks like the sawfish
problem. More on the rest is on -policy.
> goswin> Configuration using debconf should of cause be strongly
> goswin> encouraged and important messages should be saved in a logfile and
> goswin> displayed at the end all together instead inbetween unimportant
> goswin> output.
>
> Are you considering minimal downtime for critical daemon packages?
That would be one effect.
Several times I had the problems that I upgraded via network and sshd
got stoped and I got kicked out. But I wasn't stupid, so i upgraded
the system in a screen session. But then some stupid packages pops up
and asks me to press return, so sshd wil not start up again and I have
to drive over to the system.
At the moment one cannot upgrade ssh together with some other package,
because that could ask some question and stop sshd from coming back on
again. That should not happen. Same with the pcmcia utils.
Thats stuff that can actually cost money and if it does you will hate
the stupid sawfish *sorry, but thats a _absolutly random_ :) name,
nothing against sawfish, change at will* for bothering you to drive to
the system and press return.
> goswin> At the moment the interaction seems needed, but it should go
> goswin> in the future.
>
> This is a nice, fluffy statement, and seems tob e mostly
> opinion. Have you looked at installation interaction in more than a
> handful of packages?
I expressed it more clearly on -policy, but just to clarify. I
completly seperate configuration and installation here. Configuration
must be interactive, but installation must not. Thats MY opinion.
And in all Packages that I looked at I didn't see one thing that is
not configuration and needs user interaction during installation that
is not some workaround for some dirty code.
MfG
Goswin
Reply to: