Re: A few observations about systemd
Le lundi 18 juillet 2011 à 10:35 +0100, Jon Dowland a écrit :
> I've just written pretty much the opposite in my last message to the thread,
> however: it's my opinion that supporting kfreebsd et al should be done with the
> minimum impact on the Linux Debian distribution. So, pre-supposing systemd, I
> see three options:
> 1. carry portability patches against systemd locally
> 2. support multiple init systems
> 3. drop kfreebsd (and HURD and others)
> I thought 2. would be more likely than 1. (and fairer on Tolleg!) since I
> expect there will be people with no interest in kfreebsd/HURD that nevertheless
> would like init system choice; however I'm not one of those people, and I'm
> increasingly of the opinion that choice for choices sake harms us.
There is no point into being able to choose an init system. It’s better
to have one that works well than three that don’t. Worse: if you have 3
init systems you usually have to cater with the features of the less
powerful one, without fully benefiting from the better ones.
As for people who can’t get out of the 70s and complain that we change
their init system and their carefully tuned ordering schemes, they can
use an OS from the 70s.
Also consider the amount of work for daemon maintainers: you would have
to maintain both a systemd service and an old-style init script.
Evidently the one that’s less used by maintainers will just rot and
you’ll end up with lots of unexpected combinations that are merely a
source for bugs.
I agree that keeping insserv for kfreebsd is an easier way for kfreebsd
porters, but I’d prefer to see kfreebsd use the same init files as Linux
- this could be done by porting systemd, but a simpler compatibility
layer to generate old-style scripts from systemd services would also do
.''`. Josselin Mouette
: :' :