Re: Perfect Jessie is something like this...
Am 23.10.2014 00:18, schrieb Martinx - ジェームズ:
> I'm seeing that systemd is spreading itself everywhere, kind of creepy but,
> well, I think that it is better to learn about it now, instead of fighting
> it.
Note that in my eyes, you should differentiate between two different things:
- mandatory dependencies on stuff developed in the systemd package
(such as logind)
- optional and harmless dependencies that enhance software when using
systemd but don't hurt it otherwise (libsystemd0)
If I put my developer hat on, it would be my goal to integrate it as
best as possible with components used by different distributions. So,
yes, I'd include optional systemd socket activation code if that makes
sense for that software. An while I don't personally care that much
about Upstart, if people expressed enough interest in it, I'd add code
for socket activation there. And since socket activation is a runtime
feature that is harmless if not used, I wouldn't make those dependencies
optional, because they don't hurt anybody if the feature is not used.
On the other hand, adding a mandatory dependency on something,
especially if it requires some software to be running and not just
installed, is something that I would have to consider much more
carefully, because it reduces the amounts of environments my software
runs in, and not increases it. My guess is that most if not all people
developing software would agree with that sentiment. The flip side is
that at some point, there might be some software that it so useful and
without any realistic alternatives, that weighing the options I might
come to the conclusion that I might not be willing to pay the price of
not depending on it.
And essentially, that's what the GNOME developers did when they decided
to have a mandatory dependency on logind - they weighed the options and
came to the conclusion that the cost of not requiring it was too high
for them. You might disagree with their conclusion, because of a
different set of values you espouse. But realize that the GNOME
developers came to their conclusion honestly and without malice towards
you or other people. (And as I said in another mail, I disagree with a
lot of decisions of GNOME in regards to UX/UI, but I never thought that
they had bad intentions.)
At the same time, it would be nice for some people here on this list to
recognize that the systemd developers also don't have bad intentions.
Every part of the systemd package they developed tries to provide a
useful building block for the operating system. Every piece of code they
wrote either does something new, solving a problem that wasn't solved
before, or does something old in a different way than before, in a way
that the systemd developers would argue is better. You might disagree
with the design decisions they made (I've done so in the past; recently
I unsucessfully tried to argue with Lennart Poettering about the
semantics of ProtectSystem= in units on the systemd ML), you might not
like the way their project is organized, etc. But please keep in mind
that all the components that systemd provides that people want to depend
on (mainly logind) had no alternative before that made people happy.
Take the case of logind: there was ConsoleKit before, it did mostly
work, but I haven't seen many people very happy with it; a couple of
years ago there were a lot of people that wanted something better, but
nobody stepped up and actually made something better (or significantly
improved ConsoleKit), until logind came.
The main reason why systemd has become so successful and people are
depending on parts of the project is that it has become very useful, and
for the most useful parts often without a realistic alternative in
sight. It doesn't mean people have to like the outcome of it all, it
doesn't mean that critics are necessarily wrong, but they should
recognize the reason for that outcome.
Basically, if somebody wants to change the current situation, they have
two options: either write better replacements for logind/etc. themselves
(and if it's really better, projects like GNOME will probably support
it) or convince other people to do so. If one is just a user and not a
developer, they are typically left with the second option. Just please
keep in mind that constantly clogging up mailing lists with innuendos,
assumptions of bad faith on the part of other people, hatred and so
forth is not going to help that cause. I'm not saying that people should
never complain, but the amount of unproductiveness I've read on the
Debian mailing lists in this regard in the last few months is just
mind-boggling.
To quote Russ Albery[1]:
> I expect I'm not the only person who
> finds the screaming and yelling decidedly unmotivating, and who has cut
> back on Debian work as a partial reaction
My guess is that if people stopped ranting so much about systemd and
started helping those people who are actually working on stuff that
makes software work without it (take the people working on systemd-shim
in order to make logind work, who've probably not even remotely gotten
enough credit by the people who are against systemd), Debian's support
for non-systemd systems would already be in a much better shape. If
you're not a developer, you can still help by testing configurations
without systemd as PID1.
Christian
[1] https://lists.debian.org/debian-vote/2014/10/msg00402.html
Reply to: