Re: Request for Comments: Standardize enabling/disabling of system services
Hi,
On Wed, Apr 01, 2009 at 09:50:47PM +0300, Lars Wirzenius wrote:
> ke, 2009-04-01 kello 20:30 +0200, Patrick Schoenfeld kirjoitti:
> > You finished reading my mail after that paragraph, didn't you? ;)
>
> Pretty much. It looked long and complicated and I was in a hurry. I
> skimmed it but I see now I missed that you actually knew about
> policy-rc.d.
>
> Let me make amends by suggesting you _not_ have packages modify a
> hypothetical /etc/rc.conf file, and instead invent a /etc/rc.conf.d
> directory. It's been one of the lessons learned by Debian that modifying
> configuration files in a lot of maintainer scripts is error-prone,
> whereas having packages include configuration files in .d directories is
> almost foolproof.
you missunderstood my mail in this point. I explicitly stated that
maintainer scripts *must not* edit the file file. Its a file which shall
be 100% under administrator control.
> I would keep things simpler, though. Let package maintainers exercise
> their best judgement as to whether their services should be started upon
> package install or not, and provide an optional default implementation
> of policy-rc.d (i.e., something not installed by default, at least not
> until it has proven itself) that reads /etc/policy-rc.d.conf to see
> which services to allow and which not. Syntax might be something like
> this:
Basically that is the idea, just that I already stated some possible
implementation details and tended to re-use existing methods that users
are familar with.
> Deny *
> AllowPriority standard important required
> Allow apache2
>
> Default should be to allow everything. Allow sysadmin to match either
> names or package priorities.
Is dislike that format, because users are already used to the RUN_*
system and additional people changing from another distribution or even
operating system will notice similarities, which is good as well.
> This still lets packages use whatever /etc/default/$package variables
> they wish, but I'm fine with that.
Well, except for the RUN_* variable they could still do what they want.
Best Regards,
Patrick
Reply to: