[arch-general] SystemD poll

Felipe Contreras felipe.contreras at gmail.com
Fri Aug 24 19:55:16 EDT 2012


On Fri, Aug 24, 2012 at 2:47 AM, Norbert Zeh <nzeh at cs.dal.ca> wrote:
> Felipe Contreras [2012.08.23 2214 +0200]:

>> Notice that I said "probably". Again, I don't *need* to provide any
>> evidence because I'm not making the claim that systemd has problems,
>> or that it's not ready, I am simply asking for evidence that it is.
>
> I tried to keep my mouth shut but can't resist to reply here because I simply
> don't understand how you think the world works.

> Do you want to see proof that every piece of open-source software is ready to be used?  That's ridiculous.

No, I never said anything like that. All I said is that the one that
makes the claim has the burden of proof. If you don't want to carry
your burden, then we can ignore the claim, and the discussion
continues.

Q: Is systemd ready? A: We don't know.

All right. The discussion would continue along the lines of: does it
have benefits that outweigh the _potential_ breakage that it might
cause?

Well, the developers have already said they are not going to provide
an analysis of the benefits vs. disadvantages. So we just have to
trust them. IOW; blindly submit to authority. And/or chase random
comments in blog posts and mailing lists to see what individual
developers argue. So it's all very vague. If there are no problems
when systemd is made the default, then everything would be fine. But
if shit hits the fan... well, the fact there was no public analysis,
nor evidence that systemd was actually ready, is going to upset the
users that hit these problems. But hey, it's much easier to just
ignore that fact and hope everything will be fine.

But I presume you are not even willing to forgo that claim. You are
probably arguing for the sake of arguing, show that I'm stupid, or
something like that.

>  Even though I'm not a systemd fanboy and probably would have been equally happy continuing with init scripts for a while, my general impression is that systemd is in the latter category.

Perhaps it is in the latter category, but the problem is that this is
one of the essential parts of the system. You wouldn't want Arch Linux
developers to move from Linux to HURD, or glibc to uclibc, or any
other radical move that has a potential to break *everything*. Would
you?

Not all software has the same importance in the system.

> It works for a large number of people (the majority?),

Where is the evidence of that claim? Or is that just an assumption?

But I will grant that for the sake of argument. Let's suppose that it
works for the majority of people: 60% of them, or lets be more
generous: 90%. What would be the effects on the rest (e.g. 10%)? The
system would not boot. Somebody runs 'pacman -Syu', and in the next
boot, the system just stops... that would be upsetting, wouldn't it?

Simply having that as a possibility would warrant a little extra care.
Wouldn't it?

> it works flawlessly for me so far.

You are a single data-point. Who cares if it works for you?

> Yet, here you are raising doubts about systemd being ready to be
> used.  That puts *you* in a position to explain why you think there are serious
> reasons why systemd should not be adopted.

No. Let me try to explain (yet again) the concept of burden of proof.

Suppose for some reason the majority of scientists believe in the
theory of the Big Bang. And then I come along and wonder... where is
the evidence? Well, if the Big Bang theory has merits, there would be
tons of evidence, and any decent scientist that believes in this
theory would gladly point me towards that evidence. But what happens
if scientists tell me: "no, we already believe in this theory, so now
*you* have the burden of proof if you want to discredit it"; that
would be worrying. I don't want to discredit it: I'm simply a rational
person that is looking for the evidence, and as any rational person, I
would not take the scientists words for an answer (fallacy of
authority), or accept the status quo (appeal to tradition).

So you see, status quo doesn't make the burden of proof shift into
another direction. There was a time when initscripts was all there
was, and now the choice is systemd. Well, where is the rationale
behind this? Is systemd ready? This is a perfectly valid question, and
if you don't have the evidence, you can simply say "We don't know",
and that's a perfectly valid answer. But then there would be other
questions.

So I'm sorry, but no; you cannot simply shift the burden of proof.
Notice that there's a difference between a) systemd is ready, b)
systemd is *not* ready, c) we don't know if systemd is ready or not.
Both a) and b) would need evidence to be substantiated, and the
default position is c). If nobody provides evidence either way, the
only rational stance is c).

Hopefully that's clear enough.

> In law, the default assumption is that the accused is innocent and any claim to the contrary needs to be proven.

No. The default assumption is that the accused is "not guilty". And
notice that there's a difference between "not guilty" and innocent;
innocence can only be proven with evidence, but you don't need to
prove you are innocent, all you need to do is remain the position of
"not guilty" by showing the evidence for guilt is not valid.

http://www.oregoncriminalattorney.com/Criminal-Defense-Overview/Innocent-V-Not-Guilty.shtml

> You seem to assume that the default assumption is that the software is broken

No. The default position is that we don't know if the software is broken or not.

This is fine for most of the software, as if it's broken, the user
will simply avoid it after installing it. But when making it a hard
selection: for example: you cannot use KDE, only GNOME, then hopefully
you have evidence that GNOME is not broken (and incidentally that
supporting KDE would be hard work, but that's a bit beyond the
analogy).

Cheers.

-- 
Felipe Contreras


More information about the arch-general mailing list