init script bugs and changes

[Ville Skyttä]

On Thursday 05 July 2007, Jesse Keating wrote:
> Harald, I appreciate your dillagence in filing all these bugs, however I
> think a change like this needs to go through the Fedora Packaging
> Committee. You're asking folks to change content in init scripts when there
> hasn't been much / any discussion on the topic.  If we're going to make
> this policy going forward, it really needs to be in the packaging
> guidelines, and approved by FESCo.

Some questions, I'm sure there are more:

Starting an already running service and stopping an already stopped one should 
return with 0, fair enough.  Should the init script output something in these 
cases?

Are there default values for Default-Start/Default-Stop?  If I 
specify "Default-Start: 2 3 4 5", do I also need to specify "Default-Stop: 0 
1 6" or can I leave the latter out?  What happens if I do leave it out?

If there's no Default-Start, is Default-Stop needed?  Is "Default-Stop: 0 1 2 
3 4 5 6" the same as leaving Default-Stop out?

Does chkconfig (when invoked as chkconfig, not the LSB tool symlinks) process 
Default-Start and Default-Stop lines?  If both the usual chkconfig: line and 
Default-Start/Default-Stop are specified and their information differs, which 
one prevails?  Is "Default-Start: 2 3 4 5" semantically equivalent 
to "chkconfig: 2345 xx yy" ie. it means that the service is enabled by 
default?

Which keywords are required?  None/all/subset?  Note that rpmlint currently 
gives most likely wrong advice about the keywords, more discussion at 
http://rpmlint.zarb.org/cgi-bin/trac.cgi/ticket/71