Re: If Linux Is About Choice, Why Then ...
- Date: Sat, 8 Apr 2017 22:33:30 +0200
- From: <tomas@xxxxxxxxxx>
- Subject: Re: If Linux Is About Choice, Why Then ...
-----BEGIN PGP SIGNED MESSAGE-----
On Sat, Apr 08, 2017 at 11:56:10AM +0200, Nicolas George wrote:
> Le nonidi 19 germinal, an CCXXV, Martin Read a écrit :
> > If a systemd unit for a particular service needs the attention of an expert
> > in order to be robust, the SysV-style RC script for the same service
> > probably also needs the attention of an expert in order to be robust.
> > As such, I find your suggestion that declarative configuration causes
> > 'stronger separation between "priests" and "mortals"' more than a little bit
> > questionable.
> I think Tomás is perfectly aware of that, and quoted that argument from
> systemd opponents without making it his own.
I'm aware of that, but still make this argument (in part) my own. There
is a whole spectrum between a script that "works in my environment" and
a robust script, as Martin envisions, the kind you would package as part
of a distribution, having to cope with very different environments.
The beauty of that spectrum is that a "mere mortal" can walk this thing
gradually, improving in the process.
I think it's perfectly legitimate to disagree with me on that, but there
> But you raise an interesting point. The people who invoke that argument
> do not realize that they are already experts, "priests", of the shell
> scripting language.
> I think that explains some of the most vocal systemd opposition: systemd
> aims to get rid of the scoriae of the past, but since it is IMHO
> somewhat over-engineered, it has a learning curve that is rather steep
> at the beginning. People who painstakingly learned the specifics of
> shell scripts and init scripts are afraid that their skill will lose
> value or become obsolete and they will need to start again.
It's more: there's a huge gap between "doing what systemd allows", in its
declarative language, and changing the way it works, which involves
grokking the C sources.
This kind of layering is what we software "engineers" do all the time, but
in this case, the layer gap is far too wide for my taste.
- -- tomás
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
-----END PGP SIGNATURE-----