VSIDO systemd and upstart discussion

jedi

#15
OK, my .00002 cents worth. (sorry for the lengthyness of this post, it touched a nerve maybe or something close)
First, what is the problem with sysvinit?  OK, so systemd will do a complete userspace boot-up in 900ms.  So what?  Having personally tried systemd with VSIDO, I saw no 'great' benefits. OTOH, I have never tried Upstart.  (does using Upstart mean using Mir/Unity?)
Second, when you mention Upstart, all that I can think of is "we're Shuttleworth'd"...
And third, why do we have to have one, or the other?  There are multiple choices now.  (As Digit so aptly mentions above) 

Back to boot speeds, I read one posters comment, (on one of the other forums) and I'm paraphrasing here, if my system takes 3 to 5 minutes to boot, but once booted it remains solid and has no issues, then I don't care.  Personally I think that the boot speeds that most people are talking about these days have more to do, generically, with UEFI than systemd. (yes, I know systemd drastically can reduce boot times, and have seen that with my own eyes) But, UEFI passes the boot off from the BIOS (read; slow) and directly to hardware, which, if your using an SSD, can seem instantaneous.  So, in a server environment, IMO, boot speeds tend to be irrelevant.  On a desktop (laptop) this becomes even more irrelevant. (unless of course, your James Bond, the earth is getting ready to explode, and you have to save the Universe, then I'll agree, you need a fast booting piece of metal)

I also feel that for the most part, if your using Linux, and your making decisions like this (hmmmm systemd or Upstart) then most likely, you can pick and choose your init system for yourself.

The nice thing about systemd, to quote from the Debian wiki: "Fedora, OpenSuSE, Arch and Mageia have already made the choice to use systemd, and it is getting excellent upstream support for a growing number of packages".  Another nice 'feature', "The transition plan is easy, since existing init scripts are treated as first-class services: scripts can depend (using LSB headers) on units, units can depend on scripts. More than 99% of init scripts can be used without a modification". (another quote from the Debian wiki) Wow, without modification?  Nuff said 'bout dat.

Gentoo, I think, is into OpenRC, but also supports systemd.  I always kind of looked at Gentoo as the true geeks Linux!  No offense intended there...

While I ask, who would use it, doesn't Gnome3 depend on systemd?  In realistic terms I would bet that if your not using a 'buntu' flavored derivative, (unity) stats would say that A LOT of 'new' users are using Gnome3. Does this affect their decision to part ways with Linux?  Same question for KDE4...

At least 2 of the Tech-Committee members at Debian work or worked for Ubuntu/Canonical/Shuttleworth.  Does this mean an "automatic" vote for Upstart?  At least one of them has publicly stated his support for Upstart. (Ian Jackson)  This scares me.  I think there are actually too many nuances to this discussion that we'll have no say in.  If it comes right down to it, my stubborn and anti-establishment mind set automatically makes Upstart a no go for me.  While I appreciate what Ubuntu has done for Linux as a whole, (bringing in LOTS of new users, that once they find out how intuitive Linux can be, run for the Linux distro hills so-to-speak)  The commercial an PR aspect of canonical/Ubuntu really discourage me.  I feel that Shuttleworth wants Linux as his own personal kingdom similar in aspects to little Billy Gates and his desire for MS to eventually rule the world, then on tot the Universe!

I have no personal experience with OpenRC, epoch, or runit.  SysVinit, and systemd on the other hand I have plenty of personal exp with.  I have no complaints about either.  IMO systemd has a lot of dev's working hard to keep the upstream satisfied in that it already supports so much.  I think Upstart has a lot of catching up to do in this regard.

Also back to philosophy; FOSS, OpenSource are the bare bones basics of Linux to me.  I have yet to actually HAVE to pay for anything related to Linux.  How many of you have?  systemd doesn't threaten me or my stand on the FOSS project goals.

To specifics.  I, like VastOne, am concerned about Lennart's attitude.  The guy is brilliant, a genius if you must.  However his curt, and sometimes antagonistic views, are a real turn off.  Brilliance sometimes has it's idiosyncrasies/eccentricities, and I take most of his angst as par for the course with how brilliance works.  I happen to really like pulseaudio.  Lennart was the original dev there if I'm not mistaken.  The project almost died.  Since his leaving, pulseaudio has continued to improve, and I think if the same happened with systemd, the same would happen that happened with pulseaudio.  There are plenty of competent dev's working on it at this point that I'd say it is completely reliable in being maintained.  Just as pulseaudio continues to improve without him. (Lennart)  This is not meant to be offensive to Lennart, just stating what I've read and heard talking to other dev's.  To clarify, I've never spoke with Lennart and have no personal feelings there at all and am not trying to start a flame contest!

Debian's results in their systemd survey were pretty succinct;  If not overwhelming support for systemd, it was substantial.  You can't make everyone happy all the time, and when you try, YOU FAIL! Period.  There are many examples of this in the 'nix' world.  Here today, gone tomorrow.

All of this said, I don't know a lot about any of it!  Just reading I've done over the past month or so a little at a time.

OTOH, tomorrow, OpenRC may make a breakthrough and all of this dribble will be for naught.  All of a sudden we'll all want OpenRC.  After all, what's good for Gentoo must be good for everyone else, right?

statmonkey, in the end, is right of course.  This will impact every Debian user starting now for some, but certainly all others in the future.  Another important note in statmonkey's post,
Quote from: statmonkeyI believe I mentioned to him that his "truculence" was not earning him many converts and his response was pretty much "so?"
This seems to be a trend especially the last year with some of the better Linux dev's.  Alienation is the last thing needed here.

In the end, I am not touting one over the other.  However, it seems that one has been totally thought through and the other has not.

Being listed on the Canonical agreement has me worried about LightDM in a big way. I was not aware of that, and I also am quite fond of LightDM and, I suppose I'll have to start looking at a different option. Unfortunate, as I truly liked LightDM and the ease with which you could script your way to personalized setups. Oh well...

Quote from: DigitPS.   having said all that... if it aint broke...
Forum Netiquette

"No matter how smart you are you can never convince someone stupid that they are stupid."  Anonymous

Digit

i'll just get this tidbit outta the way first...
QuoteGentoo, I think, is into OpenRC, but also supports systemd.  I always kind of looked at Gentoo as the true geeks Linux!  No offense intended there...
gentoo (and presumably funtoo) can do init/bsd style stuff too, as well as epoch (i think).  likely other ways too.  exherbo lets you go openrc or systemd out of the box, both mentioned in the installation, iirc.  there's a page on the wiki or something depicting it...  ah, yup: http://wiki.gentoo.org/wiki/Comparison_of_init_systems   ... not that this should hold much bearing on sway on our deliberations for vsido.  ;)

then, y'see..
QuoteAnd third, why do we have to have one, or the other?  There are multiple choices now.
hrmmm... let the user choose, in the installer... hrmmm....  ...  i can only imagine that will involve a tremendous amount of legwork for our poor dev, to accommodate them all.  would be sweet as tits, but then the odd-balls, systemd and upstart, they'd throw spanners into the works complexifying and increasing bugs and limitations and lengthy workarounds needed... or so i presume from what i've gathered...  never really looked into the code, nor likely would make much sense of it ~ did actually skim through systemd's code when new ~ and i suppose that's why i like the idea of a fathomnable init system, shell scripts as much as is possible/sensible.     ...but i suppose that's just me and some crazy idea i have of wanting to know what's going on in my system. ;)

"shuttleworth'd"

yeah, we dont rly wanna be that, do we?  associated with a guy who's said he has taken a leaf from the book of rockerfella, gates and soros...  o_O

Quote"The transition plan is easy, since existing init scripts are treated as first-class services: scripts can depend (using LSB headers) on units, units can depend on scripts. More than 99% of init scripts can be used without a modification"
ghadamn, i didnt realise there be such a seamless transition!  :O  ... so we could be in entirely familiar territory?  ... what's the catch?  does this inhibit the speed boons (which i dont really care all that much about anyways), or add potential nasty bugs for the 1%, or...wha?  there's gotta be a catch.  i'm slowly, bit by bit, losing reasons not to jump on the systemd bandwaggon. ^_^

Quotedoesn't Gnome3 depend on systemd?
o_O  good grief!  that's absurd!  ... on gnome3's part.  ... oh well, so long as systemd doesnt depend on gnome 3, lol.
either way though, it seems like foot shooting.

ace post there jedi.

if we were given the vote, and it was a fixed choice of either systemd, or upstart, i think it's no competition at all.  though i still would have a little tear in my eye as i voted for systemd.  upstart has seemed from the start like a kind of afterthought response to system d, like "oh, that can be done, a new init system from the ground up, cool, we'll want that too for shuttleworth kingdom".

Digit

oh hey, look...

http://lwn.net/Articles/513224/ & http://lwn.net/Articles/512719/
https://wiki.debian.org/Debate/initsystem/multiple (instantly reminds me of how openrc isnt itself a whole solution, and relies on an init system, and it just takes care of managing things nicely)
https://wiki.debian.org/OpenRC ( & http://lists.debian.org/debian-devel/2012/04/msg00547.html )
https://wiki.debian.org/Debate/initsystem/openrc (has nice subsections: https://wiki.debian.org/Debate/initsystem/openrc#A_bit_of_OpenRC_description (1,2&3))


.... so i suppose i wasnt so mad after all for suggesting it be considered. ;)

ps. mmm, i do lub me the neddyseagoon and bonsaikitten. :)

PackRat

#18
@jedi - the Debian wiki lists Fedora as switching to systemd? Interesting since the Ubuntu Upstart page has them listed as using upstart since Fedora 9 (that's what, 11 releases ago?). It may be worth some research to see if Fedora tried Upstart then went systemd - based on some of the comments about servers in the links by statmonkey, I can see that decision coming down from Red Hat.

@jedi and statmonkey - not to highjack the thread, but if you're concerned about the dm you're using there is the Console Display Manager (CDM) written by an Arch user -

https://bbs.archlinux.org/viewtopic.php?id=84408

I know several users (gutterslob and myself come to mind) that have used this in the past successfully on Debian systems. Needs some initial setup, but nothing too extreme. Currently, I'm using lightdm and slim (with Slackware).
I am tired of talk that comes to nothing.
-- Chief Joseph

...the sun, the darkness, the winds are all listening to what we have to say.
-- Geronimo

statmonkey

Ha - Of course PackRat had a solution.  Thanks will look into CDM thank you.

I need to find the thread but at some point Lennart was admitting that if Debian accepted systemd it was inevitable (due to structure) that dev's would make systemd a dependency.  In fact, he suggested that whether it was systemd or upstart it would be likely in the future that due to the bundling of udev, etc. this would be a goal.  The point I am getting at here is that a decision not to use either of these is more than likely going to impact the other packages in the distribution.  I am not sure I am putting that as well as it should be but it's very critical to the whole issue. This I believe is IG's objection to the whole discussion.  Should Debian go in the upstart direction some of our favorites could no longer be available and the freedom of movement limited down the road by either system.

Some great stuff here from Jedi etc.  Digit thanks for the links.  But I have to call bullshit on that 99% don't have to be touched.  Systemd changes the method of initialization and reporting.  It blackboxes certain operations and it is not compatible with many of the old sysv stuff.  I don't know where they get that but I have 4 bugs right now that apt can't fix and that have been open since at least Jan of last year.  The threads that I have read and responded to frequently include information about patches that have been written but aren't yet available due to ancillary issues caused by their implementation and subsequent dependency issues.

Short night and lots of work today so this post might sound a bit alarmist (my apologies) but it's great that so many of you have been thinking on this. This is such a fantastic community, I really had begun to wonder if I was alone in the wilderness regarding my concern about the path Debian seems to be on.  I don't think the multiple choice method is anything that v-ger :) could do without killing him and I am not sure that the integration level of systemd will allow that long term.  upstart is just a no go for me, Jedi's post made that confirmed in my mind.

Jedi - yes Ian and one other have essentially committed publicly to upstart  :'(

VastOne

OK, been filling my head with a lot of shit..  a lot of angst out there over these things

@jedi, great post!

@Digit .. thanks for all the info, it is relevant and I lub it but I must confess, my head now hurts

Regarding Canonicals agreement and lightdm...
QuoteWith the contributor agreement chosen by Canonical, the Harmony CLA, the contributor gives Canonical a licence to use their contributions. The contributor continues to own the copyright in the contribution, with full rights to re-use, re-distribute, and continue modifying the contributed code, allowing them to also share that contribution with other projects.

Is it really an issue? ... From that I looked into CDM and it looks like it never made it past 'hey how do I package this for debian' ... all the threads look at least 3 years old and now seems irrelevant IMO

I have looked into OpenRC and it is not an easy strip and rip function to change on a debian install... I will continue to play with it

Overall, I am wondering if the best option may be to sit in the middle of this road until there is more dust that is settled and see where it goes. Right now I do not see a clear path at all from debian and until that is done it makes this an exercise in futility IMO for VSIDO... NOT for this discussion or alternatives, it just seems as if my hands are tied right now to move in any direction... I am comfortable waiting for it to happen

I would like to use systemd as a default for VSIDO but feel uncomfortable with that right now due to all that has been discussed... so I will simply ask, what would you do?
VSIDO      VSIDO Change Blog    

    I dev VSIDO

statmonkey

I have a pet phrase that can be applied to many situations including this one. "To see the tree falling is to move".  In other words if a tree is falling on your head you don't think about it, you just get out of the way.  This works in reverse as well, if there is no tree falling you really don't have to get out of the way. 

There have been a great many links posted here, some really good input and of course a really good discussion. But, based on where you are at and I think you stated that clearly.  I would do nothing but continue to look around and keep an open mind.  Keep posting things in this thread as more becomes clear, there is a lot of cruft out there right now and when it is clear, it just will be clear.

On a separate subject, yes the Canonical Agreement is really a problem but that should be in another thread.

Regarding CDM, yes it's dead but the part that hung them up is trivial to me.  I will delve into this and see what I can do with my limited abilities.  Again another thread.

My .0000000002 cents worth (have to outdo Jedi) is to do nothing I guess.  But being aware that this is a moving target and keeping a 10,000 foot view of it is a good idea.

PackRat

@vastone -

I just pointed out CDM as an alternative for statmonkey and jedi if they wanted to move from lightdm; didn't mean to imply it should be adopted by VSIDO.
I am tired of talk that comes to nothing.
-- Chief Joseph

...the sun, the darkness, the winds are all listening to what we have to say.
-- Geronimo

VastOne

^ I know... I was actually looking forward to something different and wanted needed something new to look at so CDM was exciting!  I was giving full disclosure as to why my head hurt

No worries at all my friend...  8)
VSIDO      VSIDO Change Blog    

    I dev VSIDO

jedi

#24
@Digit, nice links, did you see this one?
https://wiki.debian.org/Debate/initsystem/systemd
or this one?
https://lists.debian.org/debian-ctte/2013/12/msg00149.html

Sorry if I'm duping things here.  Both those links have some really good info concerning the pro's/con's.

@statmonkey, I agree with the tree analogy!  Also the 99% came from that first link I posted above...
@VastOne, I think your idea to stand in the middle of the road on this for a while is a sound decision.
Reading more about the CLA and LightDM, I agree that this is probably a non-issue.

After the mind spinning/whirling reading adventure of the past few days, my decision would be unhesitatingly for systemd.  The more I read about Upstart, the scarier that proposition gets to me.  They are determined to have total control of the code, from all I've read, which is totally against any FOSS philosophy I've ever studied, they are a small group and have a limited number of participants (evidently a strict guideline of Ubuntu/Canonical/Shuttleworth) where systemd is the opposite, in that they have a lot of dev's, and they are not trying to totally control the code.  I also like that the systemd code has over 14,000 lines of comments in it, or around 10% of the lines of code...
This is definitely NOT the case with Upstart.

systemd is very well documented, with man pages on practically everything.  Upstart, not so much...

I believe there are still a lot of dominoes to fall so-to-speak, and that there is plenty of time to make a decision regarding VSIDO in the future.

What I have found scariest in this little sojourn is the fact that "Debian" has allowed Canonical/Ubuntu employees to be part of their core group of developers.  To me this sounds/feels ominous and smells like there's some devious plan in place by Shuttleworth to abscond with Debian.  Shouldn't that be the opposite?  Canonical/Ubuntu allowing some "Debian" developers to be a part?  I firmly believe that Debian should be standing on it's own, as it always has, and if a distro wants to use it as it's base, so be it.  I don't feel that Canonical is, how do you say it, playing fair here?  (But hey I'm probably just being my usual paranoid self!!)  :)
Forum Netiquette

"No matter how smart you are you can never convince someone stupid that they are stupid."  Anonymous

statmonkey


VastOne

I have figured out how to remove and replace sysvinit with systemd ... so far the only thing I am seeing at all is incredible bootup and shutdown speeds and no other issues

Should I create a x64 ISO for testing... what say you?
VSIDO      VSIDO Change Blog    

    I dev VSIDO

VastOne

One other thought... Sitting in the middle of the road does have distinctive advantages.  Should debian decide to go to upstart, we can say no thank you and go with systemd if needed

Another thought I have is systemd is being 'pushed' as a redhat product... do you in general trust redhat and fedora more or less than Canonical?
VSIDO      VSIDO Change Blog    

    I dev VSIDO

zbreaker

Quote from: VastOne on January 02, 2014, 11:16:35 PM
Another thought I have is systemd is being 'pushed' as a redhat product... do you in general trust redhat and fedora more or less than Canonical?

Incredible discussion gentlemen, to which I have lurked as my expertise pales to some extent to yours, but in answer to the question I guess Redhat/Fedora would be my pick.

PackRat

If you have figured out how to replace sysvinit with systemd, then create an alpha release iso for testing - part of the learning process should you decide to stay on the systemd path.

I would definitely trust RedHat more than Canonical.
I am tired of talk that comes to nothing.
-- Chief Joseph

...the sun, the darkness, the winds are all listening to what we have to say.
-- Geronimo