tasks and hooks and scripts - time to reconsider?

[Thomas Neumann]

hello

On Friday 05 December 2014 16:54:58 Moritz Struebe wrote:
> Am 2014-12-05 um 16:40 schrieb Thomas Neumann:
>> If you want to know how something os-specific is done then it's _all_
>> in your config space. There's no more magic going on behind the scene.
 
> And this will also keep your config space from getting fixes (e.g. if
> you suddenly have to start sorting packages yourself[1]).

Disclaimer:
  I don't care about this specific bug at all. I don't use the debconf-feature
  and for SuSE/RedHat installations it was irrelevant anyway. But I'd like to
  take this chance and talk about installation related bugs.

As someone who's using FAI for some time and already having a config space:
 -> I would prefer to _not_ receive this fix. 

As someone who's new to FAI and trying to set up a new config space:
 -> I would prefer to receive this fix.

That may seem a bit odd, but my rationale is: If I have an existing config 
space then it can be assumed it's working fine as it is. Even if that means 
it's working fine despite having actual bugs. However even with these bugs one 
can successfully install a machine and it's matching his/her acceptance 
criteria. (e.g.: it boots successfully)

Automatically fixing stuff may be a good thing or it may result in sudden 
breakage. It all depends on what the config space is doing. Sometimes 2 bugs 
cancelling each other out is a good thing. Manually having to fix stuff ~may~ 
actually be a good thing in this case because you have a chance to fix _both_ 
bugs.

> The challenge is to put as much as possible into the package to keep your
> adjustments to the config to a minimum while giving a maximum of freedom.

The challenge is to let the user do what he wants to do and not break existing 
setups.

I think fixing bugs is good. But if these fixes result in suddenly installing 
zombie machines refusing to boot properly then I might get a bit grumpy. I 
don't try to reinstall all my different configurations after each minor FAI  
update and verify that they still work. I just assume what has worked half a 
year ago is still working. 

Refusing to boot is one of the more obvious failures. One very subtle scenario 
would be something like using MD5 instead of SHA-512 for /etc/shadow because a 
long forgotten debconf-setting is suddenly taking effect again. Ooops.

bye
thomas