Mailing List Archive

On Thu, Sep 18, 2014 at 7:11 AM, James <wireless@tampabay.rr.com> wrote:

> Mark David Dumlao <madumlao <at> gmail.com> writes:
> > You're the only one in this thread that's imposing on everyone
> > to produce anything. You're the only one in this thread that
> > SHOULD be producing anything. That's how open source works and
> > that's how it's supposed to work. We're not your unpaid researchers.
>
> I'm sorry, Volker pointed out that the pro systemd folks came to
> gentoo-user, waiving linux's dirty panties around. We ask a few
> simple questions, now you result to name calling?
>

There is no "gentoo-user separate from pro systemd folks". You made that
up. "pro systemd folks" have been part of gentoo user for years and years
now, and they've been harassed repeatedly with "simple" loaded questions
based on wrong assumptions for years and years now.

You know all those bits I mentioned to Volker about people getting FHS
wrong, or not bothering to read man pages, or not giving a crap what an
"init thingy" was and throwing public tantrums on it? I didn't make those
up. They're here, on this list, and I've had to wade in that crap for a few
years, and even in those threads where I only intended to give practical
advice like "if you want to load udev earlier, you could write an init
script for it..." or something to that effect. Only to be heaped by
plateful after plateful of vitriolic, _technically empty_ crap and
callbacks to Unix platitudes half the sayers don't even understand that
well.

Fact of the matter is systemd isn't "invading gentoo", it's part of it now,
and has been for quite a while. All those big changes many people have been
sore about on this list could have been turned into complete non-problems
if we took all the smart-brains time spent arguing this point to instead
write integration packages the way Canek did.
--
This email is: [ ] actionable [ ] fyi [ ] social
Response needed: [ ] yes [ ] up to you [ ] no
Time-sensitive: [ ] immediate [ ] soon [ ] none

Mark David Dumlao wrote:
> The code is out there. Freely available. Both systemd and sysvinit.
> If you wanted to measure both, you could, literally, in the time it
> took since you first posted in this thread till now you could have
> measured several times and left mean comments about whichever
> system you hated the most.

Unfortunately, the systemd guys keep screaming that systemd is faster,
and burden of proof is on the party that's claiming something. It's not
James'/Volker's responsibility to prove that systemd isn't faster.

That said, you guys need to stop flaming. If anything, it's easy to
dislike SysVInit because the init scripts it uses are piles of bash,
compared to a Systemd init script that has a handful of systemd config.

Is systemd starting to encompass too much? I think so, but who cares? If
we want an init manager that reads systemd-like files but doesn't do
anything else (hostnamectl, logging, udev, etc.), I guess we'll have to
make one.

Alec

On Wed, Sep 17, 2014 at 8:12 PM, Alec Ten Harmsel
<alec@alectenharmsel.com> wrote:
> Mark David Dumlao wrote:
>> The code is out there. Freely available. Both systemd and sysvinit.
>> If you wanted to measure both, you could, literally, in the time it
>> took since you first posted in this thread till now you could have
>> measured several times and left mean comments about whichever
>> system you hated the most.
>
> Unfortunately, the systemd guys keep screaming that systemd is faster,
> and burden of proof is on the party that's claiming something. It's not
> James'/Volker's responsibility to prove that systemd isn't faster.
>

I think Mark fully appreciates that if he wants to change your mind
he's going to have to work hard to do it.

I just don't think he really cares.

The argument about whether systemd is better/worse than sysvinit was a
debate back in 2012-2013. Just about anybody actually contributing to
distros has moved on since then. That doesn't mean that there is 100%
agreement on anything, just that at this point it seems unlikely that
things are going to change much either way on that front. A few
distros are likely to avoid systemd, and the vast majority are in the
process of adopting it.

With Gentoo you can run whatever you want for PID 1, just as you can
use whatever bootloader, kernel, syslog, etc you want. Not all the
init options have equal support - upstart isn't even in the tree and
few packages supply scripts for runit. But, nobody is going to get in
anybody's way if they want to introduce upstart, etc.

The fact is among those actually contributing to projects like openrc,
udev, eudev, and systemd everybody tends to get along just fine.
There is plenty of interest in finding common ground and collaborating
so that anybody switching from one to another can do so easily, and so
that these projects don't diverge where it isn't intended. It seems
like the heaviest fighting seems to involve folks who don't contribute
to any of these.

--
Rich

On Thu, Sep 18, 2014 at 8:54 AM, Rich Freeman <rich0@gentoo.org> wrote:

> On Wed, Sep 17, 2014 at 8:12 PM, Alec Ten Harmsel
> <alec@alectenharmsel.com> wrote:
> > Mark David Dumlao wrote:
> >> The code is out there. Freely available. Both systemd and sysvinit.
> >> If you wanted to measure both, you could, literally, in the time it
> >> took since you first posted in this thread till now you could have
> >> measured several times and left mean comments about whichever
> >> system you hated the most.
> >
> > Unfortunately, the systemd guys keep screaming that systemd is faster,
> > and burden of proof is on the party that's claiming something. It's not
> > James'/Volker's responsibility to prove that systemd isn't faster.
> >
>
> I think Mark fully appreciates that if he wants to change your mind
> he's going to have to work hard to do it.
>
> I just don't think he really cares.
>
> The argument about whether systemd is better/worse than sysvinit was a
> debate back in 2012-2013. Just about anybody actually contributing to
> distros has moved on since then. That doesn't mean that there is 100%
> agreement on anything, just that at this point it seems unlikely that
> things are going to change much either way on that front. A few
> distros are likely to avoid systemd, and the vast majority are in the
> process of adopting it.
>

Yeah Rich gets it. "systemd guys keep screaming that systemd is faster"
seems to imply that most of us give a tweet what PID1 you're running. When
we don't. Most often what happens is some news on systemd developments
comes up, people say "yay!", and other people say "you're destroying Linux
and gonna doom us all" and they act all righteous when we say "uh, what?"
like it matters to us what you're running.

Fact is if it's _you_ that seems to give a tweet about systemd speed, so
it's on _you_ to measure it, I don't really care what you think. The fact
that you think pid1's speed or resource usage might be a big deal is very
indicative on how badly informed you are in the first place.

It reminds me a lot of how some communities treat Gentoo users, asking them
to off the bat produce speed benchmarks comparing them to Arch or whatnot.
As if the Gentoo users gave a tweet about what other users run on their
machines in their own time... no, they very largely don't and there's no
good reason for them to be convincing other people about it.
--
This email is: [ ] actionable [ ] fyi [ ] social
Response needed: [ ] yes [ ] up to you [ ] no
Time-sensitive: [ ] immediate [ ] soon [ ] none

On 09/17/2014 10:40 PM, Mark David Dumlao wrote:

> Fact is if it's _you_ that seems to give a tweet about systemd speed,
> so it's on _you_ to measure it, I don't really care what you think. The
> fact that you think pid1's speed or resource usage might be a big deal
> is very indicative on how badly informed you are in the first place.

I don't care about systemd speed. I really am completely ambivalent
about PID1; I've run Upstart, I've run systemd, I've run OpenRC, and
they all work fine. All I'm saying is that a common point in the systemd
community seems to be its awesome performance (unless I'm reading the
wrong documentation and conversations), and burden of proof is on the
party making the claim.

But also, caring about speed and resource usage are important. If one of
the three PID1s I've mentioned took 30 seconds to boot my system, I
would not use it. If it took 10% of my RAM, I would not use it. Lucky
for us, all three are fast enough and have a small enough footprint that
it doesn't matter which is used.

Alec

On 17/09/14 23:43, Volker Armin Hemmann wrote:
> Am 17.09.2014 um 21:52 schrieb Canek Peláez Valdés:
>> On Wed, Sep 17, 2014 at 2:27 PM, Volker Armin Hemmann
>> <volkerarmin@googlemail.com> wrote:
>>> Am 17.09.2014 um 21:02 schrieb Canek Peláez Valdés:
>>>> On Wed, Sep 17, 2014 at 1:36 PM, Volker Armin Hemmann
>>>> <volkerarmin@googlemail.com> wrote:
>>>> [snip]
>>>>> Now you use this to advertise for systemd?
>>>>>
>>>>> Systemd fanbois are becoming more and more desperate.
>>>> So, systemd is used (or it has been announced that is going to be
>>>> used) by default in all the major distributions, is available and
>>>> working great in Gentoo, and many Gentoo users and developers use it
>>>> happily.
>>>>
>>>> So, yeah, we are *really* desperate, obviously.
>>>>
>>>> Thanks for the laugh.
>>>>
>>>> Regards.
>>> you will stop laughing when redhat&poettering abandon systemd because it
>>> is 'fundamentally broken' and must be replaced with something else.
>>>
>>> Probably as soon as everybody got used to it.
>>>
>>> And if I guess correctly, pulseaudio will be the driving force behind
>>> it. Because history loves repetition.
>> Sure Volker, whatever you say. I'm willing to bet the future stability
>> of my desktop and server machines that your doomsday-scenario will not
>> happen. Actually, I'm already betting on it.
>>
>> What are you willing to bet?
>>
>> Again, thanks for the laughs. You are a funny guy.
>>
>> Regards.
> I am not betting anything.
>
> But I want you to think about something:
>
> devfs was the best thing since sliced bread.
> As soon as everybody used it, it was broken and replaced.

There was no problem with this development.

>
> hal was the best thing since sliced bread.
> As soon as everybody used it, it was broken and abandoned.

That's untrue. HAL was responsibly replaced with UDisks.
As in, when Gentoo got rid of sys-apps/hal, we made sure everything was
ported to UDisks or that unported applications that were removed with
sys-apps/hal, had a direct replacement available.
It was a logical development, that's all.
>
> *kit?
> The same.
>
>
>

FUD.

On 18/09/14 03:12, Alec Ten Harmsel wrote:
> Mark David Dumlao wrote:
>> The code is out there. Freely available. Both systemd and sysvinit.
>> If you wanted to measure both, you could, literally, in the time it
>> took since you first posted in this thread till now you could have
>> measured several times and left mean comments about whichever
>> system you hated the most.
> Unfortunately, the systemd guys keep screaming that systemd is faster,
> and burden of proof is on the party that's claiming something. It's not
> James'/Volker's responsibility to prove that systemd isn't faster.
>
> That said, you guys need to stop flaming. If anything, it's easy to
> dislike SysVInit because the init scripts it uses are piles of bash,
> compared to a Systemd init script that has a handful of systemd config.
>
> Is systemd starting to encompass too much? I think so, but who cares? If
> we want an init manager that reads systemd-like files but doesn't do
> anything else (hostnamectl, logging, udev, etc.), I guess we'll have to
> make one.
>
> Alec
>

Notably Gentoo has never used entire SysV, only the init part, not the
/etc.d/rc.d part
So this POSIX sh script's are coming from dedicated *Gentoo* project,
which is sys-apps/openrc

Just clarifying

On 18/09/14 07:52, Samuli Suominen wrote:
> Notably Gentoo has never used entire SysV, only the init part, not the
> /etc.d/rc.d part

I meant /etc/rc.d of course. Typing error. Sorry.

On 18/09/2014 02:12, Alec Ten Harmsel wrote:
> Mark David Dumlao wrote:
>> The code is out there. Freely available. Both systemd and sysvinit.
>> If you wanted to measure both, you could, literally, in the time it
>> took since you first posted in this thread till now you could have
>> measured several times and left mean comments about whichever
>> system you hated the most.
>
> Unfortunately, the systemd guys keep screaming that systemd is faster,
> and burden of proof is on the party that's claiming something. It's not
> James'/Volker's responsibility to prove that systemd isn't faster.
>
> That said, you guys need to stop flaming. If anything, it's easy to
> dislike SysVInit because the init scripts it uses are piles of bash,
> compared to a Systemd init script that has a handful of systemd config.
>
> Is systemd starting to encompass too much? I think so, but who cares? If
> we want an init manager that reads systemd-like files but doesn't do
> anything else (hostnamectl, logging, udev, etc.), I guess we'll have to
> make one.

or trim it back. Conceptually, it shouldn't be too hard to remove those
extra services leaving only an init manager.

Reading posts over the years (I don't use systemd) most of that stuff
can be disabled by config in systemd anyway




--
Alan McKinnon
alan.mckinnon@gmail.com

On Thu, 18 Sep 2014 00:34:01 +0200, Volker Armin Hemmann wrote:

> AFAIR dcop was replaced, because of the freedesktop-gnome guys. Not
> because anything was wrong with it. And look where it got us. No
> improvement at all.

It wasn't really replaced as dbus was derived from DCOP, so it was more
of an evolution.


--
Neil Bothwick

Obscenity is the crutch of inarticulate motherfuckers.

On Thu, 18 Sep 2014 07:19:21 +0200, Alan McKinnon wrote:

> > Is systemd starting to encompass too much? I think so, but who cares?
> > If we want an init manager that reads systemd-like files but doesn't
> > do anything else (hostnamectl, logging, udev, etc.), I guess we'll
> > have to make one.
>
> or trim it back. Conceptually, it shouldn't be too hard to remove those
> extra services leaving only an init manager.
>
> Reading posts over the years (I don't use systemd) most of that stuff
> can be disabled by config in systemd anyway

A lot of it is disabled by default anyway, you have to turn it on if you
want to use it. Otherwise it's just there.


--
Neil Bothwick

- We are but packets in the internet of Life-

On 18/09/2014 10:07, Neil Bothwick wrote:
> On Thu, 18 Sep 2014 07:19:21 +0200, Alan McKinnon wrote:
>
>>> Is systemd starting to encompass too much? I think so, but who cares?
>>> If we want an init manager that reads systemd-like files but doesn't
>>> do anything else (hostnamectl, logging, udev, etc.), I guess we'll
>>> have to make one.
>>
>> or trim it back. Conceptually, it shouldn't be too hard to remove those
>> extra services leaving only an init manager.
>>
>> Reading posts over the years (I don't use systemd) most of that stuff
>> can be disabled by config in systemd anyway
>
> A lot of it is disabled by default anyway, you have to turn it on if you
> want to use it. Otherwise it's just there.



That's even better then.


I'm mildly bemused by these systemd threads - so much emotion. Me, I
don't have a dog in this fight so I can sit back and look at what's
going on.

Imagine the ISC-bind lovers going completely apeshit about unbound,
thinking named is about to go away forever. That's what this looks like.




--
Alan McKinnon
alan.mckinnon@gmail.com

On Wed, 17 Sep 2014 20:54:49 -0400, Rich Freeman wrote:

> The fact is among those actually contributing to projects like openrc,
> udev, eudev, and systemd everybody tends to get along just fine.
> There is plenty of interest in finding common ground and collaborating
> so that anybody switching from one to another can do so easily, and so
> that these projects don't diverge where it isn't intended. It seems
> like the heaviest fighting seems to involve folks who don't contribute
> to any of these.

Isn't that how it always is? :(

I'm sure Canek realised that a flamefest would result from his post, but
it's a rather sad indictment of us that there as been almost no
discussion of Linus's comments - I wonder how many of the "contributors"
to this thread even read the link Canek posted.

Personally, I like to read Linus's opinions no such matters; the are
always insightful, usually entertaining and occasionally correct :)


--
Neil Bothwick

Run with scissors. Remove mattress tags. Top post. Be a rebel.

On Wed, Sep 17, 2014 at 12:06 PM, Canek Peláez Valdés <caneko@gmail.com> wrote:
> It's an interesting read; I highly recommend it.
>

Indeed. Thanks for the link!

On Wed, Sep 17, 2014 at 5:34 PM, Volker Armin Hemmann
<volkerarmin@googlemail.com> wrote:
> Am 17.09.2014 um 23:03 schrieb Canek Peláez Valdés:
>> On Wed, Sep 17, 2014 at 3:43 PM, Volker Armin Hemmann
>> <volkerarmin@googlemail.com> wrote:
>>> Am 17.09.2014 um 21:52 schrieb Canek Peláez Valdés:
>>>> On Wed, Sep 17, 2014 at 2:27 PM, Volker Armin Hemmann
>>>> <volkerarmin@googlemail.com> wrote:
>>>>> Am 17.09.2014 um 21:02 schrieb Canek Peláez Valdés:
>>>>>> On Wed, Sep 17, 2014 at 1:36 PM, Volker Armin Hemmann
>>>>>> <volkerarmin@googlemail.com> wrote:
>>>>>> [snip]
>>>>>>> Now you use this to advertise for systemd?
>>>>>>>
>>>>>>> Systemd fanbois are becoming more and more desperate.
>>>>>> So, systemd is used (or it has been announced that is going to be
>>>>>> used) by default in all the major distributions, is available and
>>>>>> working great in Gentoo, and many Gentoo users and developers use it
>>>>>> happily.
>>>>>>
>>>>>> So, yeah, we are *really* desperate, obviously.
>>>>>>
>>>>>> Thanks for the laugh.
>>>>>>
>>>>>> Regards.
>>>>> you will stop laughing when redhat&poettering abandon systemd because it
>>>>> is 'fundamentally broken' and must be replaced with something else.
>>>>>
>>>>> Probably as soon as everybody got used to it.
>>>>>
>>>>> And if I guess correctly, pulseaudio will be the driving force behind
>>>>> it. Because history loves repetition.
>>>> Sure Volker, whatever you say. I'm willing to bet the future stability
>>>> of my desktop and server machines that your doomsday-scenario will not
>>>> happen. Actually, I'm already betting on it.
>>>>
>>>> What are you willing to bet?
>>>>
>>>> Again, thanks for the laughs. You are a funny guy.
>>>>
>>>> Regards.
>>> I am not betting anything.
>> I figured it.
>>
>>> But I want you to think about something:
>>>
>>> devfs was the best thing since sliced bread.
>>> As soon as everybody used it, it was broken and replaced.
>>>
>>> hal was the best thing since sliced bread.
>>> As soon as everybody used it, it was broken and abandoned.
>>>
>>> *kit?
>>> The same.
>> Yeah. So it happened with XFree86, aRts, esd, gnome-vfs, DCOP,
>> sendmail, and it will happen again with dbus (I'm willing to bet it
>> will be replaced, at least in Linux, with kdbus). And, BTW, it's
>> happening with SysV being replaced in Linux with systemd.
>>
>> It happens all the time. It's a good thing. And it happened for *VERY*
>> different reasons in each case. Also, the transition has been
>> sometimes somewhat difficult (HAL comes to mind), but most of the
>> times really easy: we used devfs when I switched to Gentoo more than
>> 10 years ago, and I don't remember being difficult the switch to udev.
>> XFree86 => X.org was also basically trivial.
>>
>> Of course systemd can be replaced; if something cooler gets written,
>> we'll switch to it. But given the team behind systemd, and the design
>> it has, it's gonna be very difficult.
>>
>> Using Linus words, you are making excuses. You can compare systemd to
>> HAL, but doing so only shows that you don't know the code, the design,
>> and the history behind both projects.
>>
>> Regards.
>
> there was no breakage with xfree-to-xorg. True. But hal, yes. No upower
> breakage. *kit breakage. The list is too long to ignore.

There was no really breakage; some distributions dealed with those
change without issues. Gentoo is special; we didn't had the tools to
rebuild all the required dependencies some years ago. Heck, sometimes
we didn't had the dependencies right.

> Arts was not something whole systems depended upon. And whatever
> gnome-thingy you depend upon, you are fucked, because those guys are
> infected with the same mindset. As soon as the bugs are ironed out and
> everybody is using it: abandom it for something else.

Oh, Volker. You really make me laugh with your ignorance.

> That has nothing to do with 'improvement', or 'development' it is just
> stupid.

It's improvement; it's just your bigotry against GNOME/systemd, your
small mindedness and your myopic vision that makes you not notice it.

HAL is special; it was a (IMO misguided) attempt to be "portable" to
the *BSDs and similar systems. The natural conclusion was that those
guys need to take care of themselves, and that's one of the reasons
why systemd is not portable and only works in Linux.

In all the other cases, it's evolutiion:

• gnome-vfs begat GVFS, which works great.
• static /dev begat devfs, which begat udev, which works great.
• DCOP and gconf begat dbus, which works great, and it will beget
kdbus, which *will* be greater.
• aRts and esd begat PulseAudio, which works great.
• SysV begat Upstart, and together with ideas from launchd and SMF
begat systemd, which works great.

You just don't get it, because as Rich says it you aren't really
involved with the development of these technologies. It's a continous
evolution of software, sometimes using the old code, sometimes just
taking ideas, design, or learning from mistakes.

> AFAIR dcop was replaced, because of the freedesktop-gnome guys.

Oh my god; did they put a gun on their heads? It could not possible be
that dbus is so much better, right?

> Not because anything was wrong with it. And look where it got us. No
> improvement at all.

You just keep showing your ignorance. Go to the KDE mailing lists, and
tell them to get back to DCOP, becuase it was better.

They will laugh at you. Just as I'm doing right now.

Funny, funny guy.

Regards.
--
Canek Peláez Valdés
Profesor de asignatura, Facultad de Ciencias
Universidad Nacional Autónoma de México

On 2014-09-18, Alec Ten Harmsel <alec@alectenharmsel.com> wrote:
> Mark David Dumlao wrote:
>> The code is out there. Freely available. Both systemd and sysvinit.
>> If you wanted to measure both, you could, literally, in the time it
>> took since you first posted in this thread till now you could have
>> measured several times and left mean comments about whichever
>> system you hated the most.
>
> Unfortunately, the systemd guys keep screaming that systemd is faster,
> and burden of proof is on the party that's claiming something. It's not
> James'/Volker's responsibility to prove that systemd isn't faster.

I don't understand all the hoopla about systemd being "faster".

Faster at what?

Booting?

The only Linux systems where I care about boot time are embedded
systems which are never going to have the resources needed to run
systemd. As for normal desktop machines, who cares? I only reboot
them once every month or two (when I'm bored and want to make sure
they will still boot up after updates).

My laptop(s) get booted a lot more often than desktops, but the boot
times have never been an issue.

The other thing I keep hearing from systemd proponents is stuff about
how it allows you to parallelize startup. I don't _want_ stuff
starting up in parallel -- that just makes it all the more difficult
to troubleshoot problems. I want things to start up one at a time, in
a determined order.

--
Grant Edwards grant.b.edwards Yow! The FALAFEL SANDWICH
at lands on my HEAD and I
gmail.com become a VEGETARIAN ...

On Thu, Sep 18, 2014 at 3:18 PM, Grant Edwards
<grant.b.edwards@gmail.com> wrote:
> The only Linux systems where I care about boot time are embedded
> systems which are never going to have the resources needed to run
> systemd.

How about containers? When I launch mariadb I'd prefer that it happen
in milliseconds, not tens of seconds. That includes setting up
interfaces, populating /dev, getting an ip, launching ssh, syslog,
etc, and so on, oh, and mariadb.

> The other thing I keep hearing from systemd proponents is stuff about
> how it allows you to parallelize startup. I don't _want_ stuff
> starting up in parallel -- that just makes it all the more difficult
> to troubleshoot problems. I want things to start up one at a time, in
> a determined order.

I hope you aren't running openrc then. It doesn't launch in a
predetermined order.

I will agree that you get far more race conditions than you do with
openrc even with parallel startup, since processes start much more
quickly.

--
Rich

On 19/09/14 03:18, Grant Edwards wrote:
> On 2014-09-18, Alec Ten Harmsel <alec@alectenharmsel.com> wrote:
>> Mark David Dumlao wrote:
>>> The code is out there. Freely available. Both systemd and sysvinit.
>>> If you wanted to measure both, you could, literally, in the time it
>>> took since you first posted in this thread till now you could have
>>> measured several times and left mean comments about whichever
>>> system you hated the most.
>>
>> Unfortunately, the systemd guys keep screaming that systemd is faster,
>> and burden of proof is on the party that's claiming something. It's not
>> James'/Volker's responsibility to prove that systemd isn't faster.
>
> I don't understand all the hoopla about systemd being "faster".
>
> Faster at what?
>
> Booting?
>
> The only Linux systems where I care about boot time are embedded
> systems which are never going to have the resources needed to run

systemd is targeted at cloud systems and fast booting which is where I
guess redhats focus is these days since they seem to have lost the
desktop space. The fact that systemd isn't potentially as reliable etc.
is irrelevant when you are looking at a more disposable cloud model
where fast start and short life predominate.

The problem is that systemd is being forced into areas where people
don't want it (inc. me).

BillK

On 2014-09-18, Rich Freeman <rich0@gentoo.org> wrote:
> On Thu, Sep 18, 2014 at 3:18 PM, Grant Edwards
><grant.b.edwards@gmail.com> wrote:
>> The only Linux systems where I care about boot time are embedded
>> systems which are never going to have the resources needed to run
>> systemd.
>
> How about containers? When I launch mariadb I'd prefer that it
> happen in milliseconds, not tens of seconds. That includes setting
> up interfaces, populating /dev, getting an ip, launching ssh, syslog,
> etc, and so on, oh, and mariadb.

OK, that makes sense. I've never used containers and only have a
vague understanding of what they are -- I occasionally use a VM or
two, but startup speed doesn't matter for them in my applications. I
assumed there must be _some_ application where boot up speed is
important, but I just didn't know what it would be.

>> The other thing I keep hearing from systemd proponents is stuff about
>> how it allows you to parallelize startup. I don't _want_ stuff
>> starting up in parallel -- that just makes it all the more difficult
>> to troubleshoot problems. I want things to start up one at a time, in
>> a determined order.
>
> I hope you aren't running openrc then. It doesn't launch in a
> predetermined order.

I'm am running openrc (with parallel startup disabled) on my "regular"
Gentoo systems. On my systems, the startup order seems to be
deterministic. [.I also have a bunch of "other" systems I boot on
occasion for testing apps/drivers -- they're running various distros
using whatever init system they default to.]

> I will agree that you get far more race conditions than you do with
> openrc even with parallel startup, since processes start much more
> quickly.

--
Grant Edwards grant.b.edwards Yow! It's the RINSE CYCLE!!
at They've ALL IGNORED the
gmail.com RINSE CYCLE!!

> • "There's still value in understanding the traditional UNIX "do one
> thing and do it well" model where many workflows can be done as a
> pipeline of simple tools each adding their own value, but let's face
> it, it's not how complex systems really work, and it's not how major
> applications have been working or been designed for a long time. It's
> a useful simplification, and it's still true at *some* level, but I
> think it's also clear that it doesn't really describe most of
> reality."
>

He doesn't make an actual argument why useful abstraction cannot be done
in complex systems.

On Thu, Sep 18, 2014 at 11:40 AM, Alec Ten Harmsel <alec@alectenharmsel.com>
wrote:

>
> On 09/17/2014 10:40 PM, Mark David Dumlao wrote:
>
> > Fact is if it's _you_ that seems to give a tweet about systemd speed,
> > so it's on _you_ to measure it, I don't really care what you think. The
> > fact that you think pid1's speed or resource usage might be a big deal
> > is very indicative on how badly informed you are in the first place.
>
> I don't care about systemd speed. I really am completely ambivalent
> about PID1; I've run Upstart, I've run systemd, I've run OpenRC, and
> they all work fine. All I'm saying is that a common point in the systemd
> community seems to be its awesome performance (unless I'm reading the
> wrong documentation and conversations), and burden of proof is on the
> party making the claim.
>

The thing is, that's a strawman. Volker is outright delusional about
systemd people breaking into his threads and forcefeeding him Lennart facts
like "systemd is faster". It's the exact opposite. Every time a systemd
thread comes up, here come the anti-fanboys whining about "well why should
_i_ use it? because it's _faster_?" as if we gave a crap that he did.

The burden of proof is on the party making the claim, but almost nobody is
making the claim -to him-. The fact that he thinks systemd's speed is
important already betrays how biased and narrow his thinking is on the
topic. Most people don't even bother with bootup speeds that cut a few
seconds off. Heck I tried to tweak my boot process with systemd and I had a
hard time getting _even_ with Ubuntu. Generally we care more about the fact
that services have actual dependencies, are written declaratively, can be
executed exactly as upstream recommends, don't have magic code hacks, are
automatically cgrouped and thus have all child processes guaranteed killed
on service down, that logs and STDOUT are tracked and searchable in the
journal, etc etc etc. Every single one of those matters more than bootup
speed, but yeah, we heard somewhere that you can tweak parallel boots to be
faster or something.

Point is he's trying to paint the picture that systemd folks rattle on and
on about its speed, but they don't. And now _we_ have to prove it? Most of
the time we're not even the ones making the claim. It's like a McD fanboy
asking a BK fan to prove that their burgers are healthier than Big Macs...
might be true, might be false, heck either company probably has info
confirming it, but it's probably the last thing on the BK fan's mind and
he's confused that it's even ever brought up.
--
This email is: [ ] actionable [ ] fyi [ ] social
Response needed: [ ] yes [ ] up to you [ ] no
Time-sensitive: [ ] immediate [ ] soon [ ] none

On Fri, Sep 19, 2014 at 3:18 AM, Grant Edwards <grant.b.edwards@gmail.com>
wrote:

> On 2014-09-18, Alec Ten Harmsel <alec@alectenharmsel.com> wrote:
> > Mark David Dumlao wrote:
> >> The code is out there. Freely available. Both systemd and sysvinit.
> >> If you wanted to measure both, you could, literally, in the time it
> >> took since you first posted in this thread till now you could have
> >> measured several times and left mean comments about whichever
> >> system you hated the most.
> >
> > Unfortunately, the systemd guys keep screaming that systemd is faster,
> > and burden of proof is on the party that's claiming something. It's not
> > James'/Volker's responsibility to prove that systemd isn't faster.
>
> I don't understand all the hoopla about systemd being "faster".
>
> Faster at what?
>
> Booting?
>
> The only Linux systems where I care about boot time are embedded
> systems which are never going to have the resources needed to run
> systemd.


You are mistaken. I've helped a friend debug problems on a couple devices
running a custom Arch system with systemd.
--
This email is: [ ] actionable [ ] fyi [ ] social
Response needed: [ ] yes [ ] up to you [ ] no
Time-sensitive: [ ] immediate [ ] soon [ ] none

Am 18.09.2014 um 01:24 schrieb Mark David Dumlao:
> On Thu, Sep 18, 2014 at 7:11 AM, James <wireless@tampabay.rr.com
> <mailto:wireless@tampabay.rr.com>> wrote:
>
> Mark David Dumlao <madumlao <at> gmail.com <http://gmail.com>> writes:
> > You're the only one in this thread that's imposing on everyone
> > to produce anything. You're the only one in this thread that
> > SHOULD be producing anything. That's how open source works and
> > that's how it's supposed to work. We're not your unpaid researchers.
>
> I'm sorry, Volker pointed out that the pro systemd folks came to
> gentoo-user, waiving linux's dirty panties around. We ask a few
> simple questions, now you result to name calling?
>
>
> There is no "gentoo-user separate from pro systemd folks". You made
> that up. "pro systemd folks" have been part of gentoo user for years
> and years now, and they've been harassed repeatedly with "simple"
> loaded questions based on wrong assumptions for years and years now.

and that makes it fine to constantly spread pro-systemd propaganga?

So.. according to your logic, it would be fine to subscribe to systemd
mailing lists and constantly post why distri X or application Y is the
best of all?

Am 20.09.2014 um 16:08 schrieb Mark David Dumlao:
> On Thu, Sep 18, 2014 at 11:40 AM, Alec Ten Harmsel
> <alec@alectenharmsel.com <mailto:alec@alectenharmsel.com>> wrote:
>
>
> On 09/17/2014 10:40 PM, Mark David Dumlao wrote:
>
> > Fact is if it's _you_ that seems to give a tweet about systemd
> speed,
> > so it's on _you_ to measure it, I don't really care what you
> think. The
> > fact that you think pid1's speed or resource usage might be a
> big deal
> > is very indicative on how badly informed you are in the first place.
>
> I don't care about systemd speed. I really am completely ambivalent
> about PID1; I've run Upstart, I've run systemd, I've run OpenRC, and
> they all work fine. All I'm saying is that a common point in the
> systemd
> community seems to be its awesome performance (unless I'm reading the
> wrong documentation and conversations), and burden of proof is on the
> party making the claim.
>
>
> The thing is, that's a strawman. Volker is outright delusional about
> systemd people breaking into his threads and forcefeeding him Lennart
> facts like "systemd is faster". It's the exact opposite. Every time a
> systemd thread comes up, here come the anti-fanboys whining about
> "well why should _i_ use it? because it's _faster_?" as if we gave a
> crap that he did.

I am deluded? Who again posted systemd propaganda again?
>
> The burden of proof is on the party making the claim, but almost
> nobody is making the claim -to him-.

No, just on public mailing lists and fora.

True, speed is not a factor.

Except if you claim it is.

> The fact that he thinks systemd's speed is important already betrays
> how biased and narrow his thinking is on the topic. Most people don't
> even bother with bootup speeds that cut a few seconds off. Heck I
> tried to tweak my boot process with systemd and I had a hard time
> getting _even_ with Ubuntu.

so the systemd-fanbois that always masturbate about how systemd is so
much faster than anything else are actually lying?

Interesting.

If those systemd-fanbois wouldn't talk about how-fast-their-toy-is, I
wouldn't care about it. I only boot to replace kernels. I don't care
about boot time, as long as it stays under 5 minutes.

> Generally we care more about the fact that services have actual
> dependencies, are written declaratively, can be executed exactly as
> upstream recommends, don't have magic code hacks, are automatically
> cgrouped and thus have all child processes guaranteed killed on
> service down, that logs and STDOUT are tracked and searchable in the
> journal, etc etc etc. Every single one of those matters more than
> bootup speed, but yeah, we heard somewhere that you can tweak parallel
> boots to be faster or something.

and if your system breaks and systemd stops working - how do you easily
access those logs? Just a question. With other logging solutions it is
easy: cat, less tail... etc.

>
> Point is he's trying to paint the picture that systemd folks rattle on
> and on about its speed, but they don't.

except when they do.

On Sun, Sep 21, 2014 at 12:28 AM, Volker Armin Hemmann <
volkerarmin@googlemail.com> wrote:

>
> I am deluded? Who again posted systemd propaganda again?
>


> Point is he's trying to paint the picture that systemd folks rattle on and
> on about its speed, but they don't.
>
>
> except when they do.
>
>
The first person who even brought up systemd's speed was making an
anti-systemd remark. Several times in this thread the need to even discuss
speed was dismissed because
very few people cared much for speed in the first place.

And the fact of the matter is that's how most systemd threads run in this
list. "systemd has a new feature" or "help me get this thing to work" or
"has anyone tested blabla yet" all invariably end up with very few
pro-systemd people even bringing up speed and many anti-systemd people
demanding that they do.

Case in point, you and your bullshit here.

--
This email is: [ ] actionable [ ] fyi [ ] social
Response needed: [ ] yes [ ] up to you [ ] no
Time-sensitive: [ ] immediate [ ] soon [ ] none