[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: systemd

On 8/7/12, Tom H <tomh0665@gmail.com> wrote:
> On Fri, Aug 3, 2012 at 3:28 AM, Joel Rees <joel.rees@gmail.com> wrote:
>> On 8/1/12, Tom H <tomh0665@gmail.com> wrote:
>>> On Mon, Jul 30, 2012 at 8:36 PM, Joel Rees <joel.rees@gmail.com> wrote:
>>>> On 7/30/12, Tom H <tomh0665@gmail.com> wrote:
>>>>> On Wed, Jul 25, 2012 at 11:26 AM, Ralf Mardorf
>>>>> <ralf.mardorf@alice-dsl.net> wrote:
>>>> I don't know what brings Tom H here, but I am becoming more active
>>>> here precisely because of systemd. Well, that, and SELinux. Too many
>>>> changes too fast, too many of them shifting from stable techniques
>>>> known to work in *nix environments to experimental implementations of
>>>> techniques that are known to be primary underlying factors of many of
>>>> the technical issues in the monoculture/world-domination OS.
>>> I have absolutely no idea why I'm subscribed and post to d-u! ;)
>> That seems odd to me.
> It was a sarcastic response to your question about my purpose in
> subscribing to this list.

Really? Blow me over with a feather.

Not that I recall asking about your reasons for being here.

I did note that I don't know why you are here, but I don't recall
asking what your reasons are.

>>> If you have genuine technical reasons for being wary of systemd, fine!
>> Uh, huh, and that's why I'm offering the heads-up.
> And the *technical* heads-up was?

Sorry you missed it. You are welcome to go back and read the archives.
Well, it turns out I've talked about one of the fundamental problems
some more, below.

And I wonder whether you will continue to insist that what I'm
bringing up is a non-technical issue.

>>> If you have an emotional negative reaction to Lennart Poettering, then
>>> your "carefulness" is silly.
>> I notice there is a certain class of politicking where the "Don't be
>> emotional!" argument gets trotted out really quickly.
> It's not "politicking!"

Yes it is. You are promoting a package. You are promoting it against a
certain amount of resistance. You are implicitly asking for people to
join the fun of testing that package. You are encouraging the
package's adoption.

That's politicking in the most real sense.

Politics is a necessary part of any organization, and so is politicking.

Being quick to brand every argument against your position which isn't
couched in strictly technical terms as emotional, well, that's a
certain class of politicking, and I pointed that out.

> All this noise about Lennart Poettering being
> this or that is silly and immaterial.

Which is why I'm willing to set aside my observations about the guy
for the moment.

> The only worthy questions are
> "does systemd work as advertised" and "if yes, should Debian adopt
> it".

Well, no, there are a lot of questions that should be asked between
those two questions, but, yeah, if systemd were structurally able to
replace sysinitv, then it would be worth asking those questions (all
of them) without reference to where it came from.

> The answer to the first is "yes,", even by Debian boot developers who
> consider sysvinit no longer up to the task, at the very least of early
> boot.

What does "even" mean?

Other than that, it sounds like you are saying that there is a group
of debian boot developers, who don't really understand sysinitv very
well, who would jump at the chance to replace it. Perhaps without
looking before they leap.

And it looks like you are trying to promote that group's opinion as
the only valid opinion. (Ergo, a certain type of politicking.)

Perhaps it sounds that way to me because of my overall technical
impression of systemv, which, I'll have to admit, is somewhat poisoned
by the confluence of systemd with Gnome 3 and SELinux. The confluence
does tend to make all certain warts in all three of those stand out in
stark relief.

> The answer to the second is "the jury's still out." AFAIR, Roger Leigh
> phrased this in this thread in terms of determining whether adopting
> systemd is in Debian's best interest.

I didn't think they had even sent it to the jury yet, really. That is,
I had the impression that evidence is still being gathered.

>> Poettering actually seems like he'd be a fun guy to hang around with.
>> If you're on his side.
> I couldn't care less.

So what is the purpose of talking about him?

>>> The reiserfs developer's a convicted murderer, that didn't make
>>> reiserfs technically inferior overnight. (Please note that I've
>>> never used reiserfs, so I don't care either way.)
>> And the primary vision which guided development of the file system is,
>> well, no longer able to be applied to the project.
>> But I've never accused Poettering of being any kind of criminal.
>> Excessive pride in one's creations and inability to deal with
>> criticism are personal failings, but they are not crimes. Not
>> particularly rare, either. (Yeah, I have personal failings, too, and
>> they do impact my projects.)
> I never implied that you or anyone else has accused Lennart Poettering
> of begin a criminal. I was simply making the point that reiserfs is
> considered to be technically very good and yet we could bad things
> about its main developer.

Very good for some purposes, not necessarily for everyone.

Developers all have their personal quirks. Some of those quirks do
impact their work. Why should we be arguing about this?

>> Yeah, talking about people tends to draw attention away from technical
>> issues, which is why I'm biting my tongue here.
> What does "biting my tongue" mean? If you have an issue with someone,
> especially someone who doesn't subscribe to this list, feel free to
> email them directly and spare us the noise. Thanks.

That's precisely what biting one's tongue means. I have issues with
the guy. I don't particularly want to discuss them here, but, as I
said way up in the thread, I could see why some people might. I have
said so, but I am not discussing those issues here.

>> The technical issues start out with this problem:
>> systemd is far too disruptive to have been stuffed into the main
>> branch of Fedora the way it was. Good engineering would have been to
>> have set up two concurrent forks of Fedora, to keep reference points
>> available when working on the integration, and to provide a safety net
>> when systemd's basic design errors surface.
> Fedora and Ubuntu adopt packages early; so what? That doesn't mean
> that those packages are bad.

systemd is not just another package. It effects everything else,
through the configuration files. In some ways, bringing it in as a
services manager is more disruptive than bringing the freeBSD kernel
in. In other ways,. maybe not quite so disruptive, but it warrants a
parallel track, just like the the freeBSD kernel does.

There is a difference, of course. The systemd fork would be expected
to either merged back in, once stable, or abandoned, if the design
issues can't be worked around, where Debian/kFreeBSD is expected to
have a parallel life of its own.

That's the way things should have been done at Fedora (given the
advertised goals). If debian developers are serious about it, it's the
way things should be done in debian.

It's the way this sort of thing should be done in any distribution
with enough developers and adventurous users to handle such a fork.
Smaller distributions with less baggage, less momentum, might
reasonably choose to be adventurous with the entire distribution, or
might reasonably choose to be conservative.

The way it's working right now, the systemd developers community seems
almost to be trying to hijack the entire Linux community and push all
the distributions down the adventurous path, RIGHT NOW, BEFORE IT'S

(And it is my impression that they are, indeed, shouting about it,
which is one of the reasons some people will find it hard to talk
about systemd without talking about the developers.)

The shouting is not technical, but the management issues are.

The management issues are not the only technical issues I see in
systemd, but I have already wasted more time on this discussion than I
had (again) today. Nearly as I can tell, the debian developers are
being much more conservative about systemd than Fedora, and that's
enough that I can afford to work with debian for the next little
while, until the dust settles, or until I can make enough time to set
up an openBSD working environment that I can work in.

Joel Rees

Reply to: