Since this list seems prepared to stick to logic, facts and solutions, rather than heat and emotion, perhaps some kind soul would answer a question for me.
I have no opinion on init systems. I accept what I am given and am grateful, but ....
If sysvinit is as staggerigly marvellous as it is being painted by some, and if it was such a near thing that it got dropped as the Debian default init system, how come the vote was apparently between Upstart and systemd? And given that it was a vote between Upstart and systemd, why are some people so up in arms and ranting that sysvinit was dropped by a meaningless margin? (The chairman's casting vote.) When and how was the decison to drop sysvinit as the Debian default init system actually taken?
Lisi
From: lisi.reisz@gmail.com To: trinity-users@lists.pearsoncomputing.net Date: Thu, 18 Sep 2014 16:17:02 +0100 Subject: [trinity-users] systemd and sysvinit, a question
Since this list seems prepared to stick to logic, facts and solutions, rather than heat and emotion, perhaps some kind soul would answer a question for me.
I have no opinion on init systems. I accept what I am given and am grateful, but ....
If sysvinit is as staggerigly marvellous as it is being painted by some, and if it was such a near thing that it got dropped as the Debian default init system, how come the vote was apparently between Upstart and systemd? And given that it was a vote between Upstart and systemd, why are some people so up in arms and ranting that sysvinit was dropped by a meaningless margin? (The chairman's casting vote.) When and how was the decison to drop sysvinit as the Debian default init system actually taken?
Lisi
Hi Lisi,
As a linux user, I'd say that the init system, as long as it works as it should and it gets you to your TDE desktop, is not a big matter to the end-user. Some years ago, changes has been made to accelerate linux boot time. As I understand it, previously, linux booted up in a very linear dos-like manner, which is one thing after the other. Changes has been made to launch different services and parts of the system in a more parallel way, to save some time. Due to these changes, classic init system were changed for different one. We don't always remember it, but, as an example, PCLinuxOS 2009 (and older releases too) took much longer to boot than the ''current'' 2010 and newer releases, due to changes in the organisation of the system bootup. Also, do you remember how linux was almost as long to boot than to shut down?
In the end, it doesn't have much negative impact on the end users.
-Alexandre
On Thu, Sep 18, 2014 at 11:49:11AM -0400, Alexandre wrote:
As a linux user, I'd say that the init system, as long as it works as it should and it gets you to your TDE desktop, is not a big matter to the end-user.
There's the rub -- does systemd work *as it should*? I don't think so.
I'm not an expert, but the systems administrators I work with have nothing but scorn for systemd. They're not all old Linux grey-beards either, in fact the ones who hate systemd the most are the young guys.
Also, do you remember how linux was almost as long to boot than to shut down?
*shrug* I don't care too much. It would be nice if my Raspberry Pi would boot up a bit quicker, but my other Linux systems -- a server and two desktops -- are on virtually 24/7 for months at a time. I don't care if they take a couple of minutes to boot up, because I only do so maybe once every six months. (And if I had a UPS at home, it would be less than that.)
As an end-user, what really concerns me about systemd that it apparently turns it into Windows: every software update requires a reboot. If true, that is, frankly, disgraceful and inexcusable. Windows has 10 or 15 years of history locking them into that behaviour. What's systemd's excuse?
On Thursday 18 September 2014 17:04:26 Steven D'Aprano wrote:
There's the rub -- does systemd work *as it should*? I don't think so.
I'm not an expert, but the systems administrators I work with have nothing but scorn for systemd. They're not all old Linux grey-beards either, in fact the ones who hate systemd the most are the young guys.
So sysadmins don't like systemd. But can you not answer my questions?
Lisi
On Thu, Sep 18, 2014 at 05:22:22PM +0100, Lisi Reisz wrote:
On Thursday 18 September 2014 17:04:26 Steven D'Aprano wrote:
There's the rub -- does systemd work *as it should*? I don't think so.
I'm not an expert, but the systems administrators I work with have nothing but scorn for systemd. They're not all old Linux grey-beards either, in fact the ones who hate systemd the most are the young guys.
So sysadmins don't like systemd. But can you not answer my questions?
Sorry, I don't know.
I'm on leave from work, otherwise I would ask the guys I work with if they know, or have an opinion. (I'm sure they'll have opinions, they are very opinionated. But whether they are *valid* opinions is another story.)
On Thursday 18 September 2014 17:28:05 Steven D'Aprano wrote:
On Thu, Sep 18, 2014 at 05:22:22PM +0100, Lisi Reisz wrote:
On Thursday 18 September 2014 17:04:26 Steven D'Aprano wrote:
There's the rub -- does systemd work *as it should*? I don't think so.
I'm not an expert, but the systems administrators I work with have nothing but scorn for systemd. They're not all old Linux grey-beards either, in fact the ones who hate systemd the most are the young guys.
So sysadmins don't like systemd. But can you not answer my questions?
Sorry, I don't know.
I'm on leave from work, otherwise I would ask the guys I work with if they know, or have an opinion. (I'm sure they'll have opinions, they are very opinionated. But whether they are *valid* opinions is another story.)
Thanks for replying. I think most people don't know - I don't. But I would like to understand what's going on a bit better under/in all the heat.
Lisi
On Thu September 18 2014 08:17:02 Lisi Reisz wrote:
If sysvinit is as staggerigly marvellous as it is being painted by some, and if it was such a near thing that it got dropped as the Debian default init system, how come the vote was apparently between Upstart and systemd? And given that it was a vote between Upstart and systemd, why are some people so up in arms and ranting that sysvinit was dropped by a meaningless margin? (The chairman's casting vote.) When and how was the decison to drop sysvinit as the Debian default init system actually taken?
I believe I read every word of the Debian TC discussion as it happened.
They looked at promised features - in particular the ability to stop runaway daemons at times other than reboot and shutdown. But that is an extraordinarily rare occurrence which I have yet to see in 32 years of professional *nix systems programming and systems administration.
They also considered that the pros of declarative startup outweighed the cons, although these directly result in systemd weaknesses such as its inability to boot with keyscripts. OpenRC offers an alternative and promising approach but was given little consideration.
The TC paid little or no attention to the real problem with systemd (and to a lesser extent with upstart) - they are power plays to force other distros to resemble RHEL (or Ubuntu).
Making your competitors copy you allows you to weaken them by making them waste effort, and allows you to market your distro as the one true Linux distro which all others aspire to copy.
It is unclear which if any of the TC were improperly influenced by Redhat (or Canonical) and which if any merely looked at the promised features without considering the more fundamental issues.
Gnu/Linux probably does need a new init, and something along the lines of the core 1% of systemd with optional cgroups support would be a good approach. But allowing Redhat (or Canonical or anybody else) to churn your distro at will is not a smart move, and thus far only Slackware and Gentoo among the majors seem to have the common-sense to resist systemd's glitter.
In the case of Debian, the TC's decision was particularly unwise because Debian supports multiple OSs and systemd only runs on Linux.
sysvinit is excellent but not necessarily "staggeringly marvelous". However compared to a distro committing suicide by systemd (or upstart) it is the preferred choice until something newer and better can be adopted.
I thank Tim and Slávek and others for making TDE optionally work with systemd without depending upon systemd - unlike Debian where systemd proponents are frantically changing packages to unnecessarily require systemd.
Mike Bird
On Thursday 18 September 2014 17:21:49 Mike Bird wrote:
sysvinit is excellent but not necessarily "staggeringly marvelous". However compared to a distro committing suicide by systemd (or upstart) it is the preferred choice until something newer and better can be adopted.
So why did Debian dismiss it before the discussion ever got to TC? Question, not rhetorical retort. If systemd and Upstart are both undesirable, and the TC only considered Upstart and systemd, then the narrowness or otherwise of the voting margin is irrelevant.
Lisi
On Thu September 18 2014 09:26:36 Lisi Reisz wrote:
So why did Debian dismiss it before the discussion ever got to TC? Question, not rhetorical retort. If systemd and Upstart are both undesirable, and the TC only considered Upstart and systemd, then the narrowness or otherwise of the voting margin is irrelevant.
There are benefits to declarative startup - although systemd does it badly and things like keyscripts don't work.
I expect sysvinit will eventually be replaced by something actually better, but neither systemd nor upstart is it.
The Debian TC may be great coders but whether through bias or naïveté they failed to consider the implications of allowing Redhat (or Canonical) to churn Debian.
I'm not sure this discussion is really on-topic for this list, though, as TDE (unlike Debian) has done it right - supporting systemd where available but not depending on it.
Mike Bird
On Thu, Sep 18, 2014 at 12:21 PM, Mike Bird mgb-trinity@yosemite.net wrote:
Gnu/Linux probably does need a new init, and something along the lines of the core 1% of systemd with optional cgroups support would be a good approach.
Just like X, sysvinit was proclaimed in need of replacement many years ago, for many reasons.
The problem is that the functionality of these packages has been built over time to answer particular problems. Any new system is going to have to go through exactly the same process, the slow evolution of "Gee, I didn't think anyone used it for that."
Sysvinit allowed for parallel boot without changing "everything". And it's especially good at being able to change pretty much any single component without a reboot.
This kind of flexibility will, I believe, turn out to be more important than the systemd developers believe it to be.
I will echo that, being just a user, I am happy to use whatever works.
I thank Tim and Slávek and others for making TDE optionally work with systemd without depending upon systemd - unlike Debian where systemd proponents are frantically changing packages to unnecessarily require systemd.
Indeed. Thank you.
Curt-
Am Donnerstag, 18. September 2014 schrieb Curt Howland:
On Thu, Sep 18, 2014 at 12:21 PM, Mike Bird mgb-trinity@yosemite.net wrote:
Gnu/Linux probably does need a new init, and something along the lines of the core 1% of systemd with optional cgroups support would be a good approach.
Just like X, sysvinit was proclaimed in need of replacement many years ago, for many reasons.
The problem is that the functionality of these packages has been built over time to answer particular problems. Any new system is going to have to go through exactly the same process, the slow evolution of "Gee, I didn't think anyone used it for that."
Sysvinit allowed for parallel boot without changing "everything". And it's especially good at being able to change pretty much any single component without a reboot.
This kind of flexibility will, I believe, turn out to be more important than the systemd developers believe it to be.
I will echo that, being just a user, I am happy to use whatever works.
I thank Tim and Slávek and others for making TDE optionally work with systemd without depending upon systemd - unlike Debian where systemd proponents are frantically changing packages to unnecessarily require systemd.
Indeed. Thank you.
Curt-
Just some system-rafiness from debian jessie:
Logfiles are binary. So you can't simply boot with a live disto and look at the logfiles. Better even, there are no persistent logfiles by default. Well, who would care to look at logfiles, anyway?
systemd/logind/journald/networkd/usersession share PID 1. Great, if somthing there goes haywire you have to reboot the system, 'cause you cannot kill one of these processes individually. On the other hand, who would ever have a system running more than a week, now that it's booting so fast? Well, and for the Windows user we have implemented the "reboot after upgrade"-feature at last. Now isn't that great?
Maybe these points are of no value for desktop users, but it's essential in my business that systems run reliably and can quite well be fixed remotely. That's not the case with systemd any more. It's diametrically to unix philosophy. It's more like "One Ring to rule them all, One Ring to find them, One Ring to bring them all and in the darkness bind them" than "Freedom of choice"
If systemd will become mandatory on Debian I already see myself packing things up and move to an other Unix land. Well, hopefully TDE will work on FreeBSD or OpenBSD then :-)
Nik
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA224
Am Donnerstag, 18. September 2014 schrieb Curt Howland:
On Thu, Sep 18, 2014 at 12:21 PM, Mike Bird mgb-trinity@yosemite.net wrote:
Gnu/Linux probably does need a new init, and something along the lines of the core 1% of systemd with optional cgroups support would be a good approach.
Just like X, sysvinit was proclaimed in need of replacement many years ago, for many reasons.
The problem is that the functionality of these packages has been built over time to answer particular problems. Any new system is going to have to go through exactly the same process, the slow evolution of "Gee, I didn't think anyone used it for that."
Sysvinit allowed for parallel boot without changing "everything". And it's especially good at being able to change pretty much any single component without a reboot.
This kind of flexibility will, I believe, turn out to be more important than the systemd developers believe it to be.
I will echo that, being just a user, I am happy to use whatever works.
I thank Tim and Slávek and others for making TDE optionally work with systemd without depending upon systemd - unlike Debian where systemd proponents are frantically changing packages to unnecessarily require systemd.
Indeed. Thank you.
Curt-
Just some system-rafiness from debian jessie:
Logfiles are binary. So you can't simply boot with a live disto and look at the logfiles. Better even, there are no persistent logfiles by default. Well, who would care to look at logfiles, anyway?
systemd/logind/journald/networkd/usersession share PID 1. Great, if somthing there goes haywire you have to reboot the system, 'cause you cannot kill one of these processes individually. On the other hand, who would ever have a system running more than a week, now that it's booting so fast? Well, and for the Windows user we have implemented the "reboot after upgrade"-feature at last. Now isn't that great?
Maybe these points are of no value for desktop users, but it's essential in my business that systems run reliably and can quite well be fixed remotely. That's not the case with systemd any more. It's diametrically to unix philosophy. It's more like "One Ring to rule them all, One Ring to find them, One Ring to bring them all and in the darkness bind them" than "Freedom of choice"
If systemd will become mandatory on Debian I already see myself packing things up and move to an other Unix land. Well, hopefully TDE will work on FreeBSD or OpenBSD then :-)
Nik
I don't see how they could get away with this on servers; uptime is typically measured in hundreds of days here and is only interrupted when a kernel update is required (e.g. when upgrading to another Debian/Ubuntu version).
TDE will not intentionally introduce a hard dependency on systemd; personally I hate the DBUS design and implementation and have avoided it wherever possible. Even the new hardware library can have all DBUS-related code disabled (albeit with some loss in functionality).
Tim
Lisi, the Debian vote, as far as I am aware, actually included all currently available options. In other words upstart and systemd were not the only choices available. The thing is 4 TC members (Canonical employees) voted for what they know, the other 4 voted for systemd (I personally think it was a deliberate vote against Canonical control) and that then left the chairmans vote which, as we know, went systemd (I also believe Garbee voted systemd to vote against Canonical control). I have read the discussion and I will try to find the actual vote, in order to see if my recollection is correct, and I will reply again if I am wrong to clarify things.
On 19 September 2014 06:38, Timothy Pearson kb9vqf@pearsoncomputing.net wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA224
Am Donnerstag, 18. September 2014 schrieb Curt Howland:
On Thu, Sep 18, 2014 at 12:21 PM, Mike Bird mgb-trinity@yosemite.net wrote:
Gnu/Linux probably does need a new init, and something along the lines of the core 1% of systemd with optional cgroups support would be a good approach.
Just like X, sysvinit was proclaimed in need of replacement many years ago, for many reasons.
The problem is that the functionality of these packages has been built over time to answer particular problems. Any new system is going to have to go through exactly the same process, the slow evolution of "Gee, I didn't think anyone used it for that."
Sysvinit allowed for parallel boot without changing "everything". And it's especially good at being able to change pretty much any single component without a reboot.
This kind of flexibility will, I believe, turn out to be more important than the systemd developers believe it to be.
I will echo that, being just a user, I am happy to use whatever works.
I thank Tim and Slávek and others for making TDE optionally work with systemd without depending upon systemd - unlike Debian where systemd proponents are frantically changing packages to unnecessarily require systemd.
Indeed. Thank you.
Curt-
Just some system-rafiness from debian jessie:
Logfiles are binary. So you can't simply boot with a live disto and look at the logfiles. Better even, there are no persistent logfiles by
default.
Well, who would care to look at logfiles, anyway?
systemd/logind/journald/networkd/usersession share PID 1. Great, if somthing there goes haywire you have to reboot the system, 'cause you cannot kill one of these processes individually. On the other hand, who would ever have a system running more than a week, now that it's booting so fast? Well, and for the Windows user we have implemented the "reboot after upgrade"-feature at last. Now isn't that great?
Maybe these points are of no value for desktop users, but it's essential in my business that systems run reliably and can quite well be fixed remotely. That's not the case with systemd any more. It's diametrically
to
unix philosophy. It's more like "One Ring to rule them all, One Ring to find them, One Ring to bring them all and in the darkness bind them" than "Freedom of choice"
If systemd will become mandatory on Debian I already see myself packing things up and move to an other Unix land. Well, hopefully TDE will work
on
FreeBSD or OpenBSD then :-)
Nik
I don't see how they could get away with this on servers; uptime is typically measured in hundreds of days here and is only interrupted when a kernel update is required (e.g. when upgrading to another Debian/Ubuntu version).
TDE will not intentionally introduce a hard dependency on systemd; personally I hate the DBUS design and implementation and have avoided it wherever possible. Even the new hardware library can have all DBUS-related code disabled (albeit with some loss in functionality).
Tim -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.9 (GNU/Linux)
iFYEARELAAYFAlQbQrsACgkQLaxZSoRZrGFV1ADfbno97D6EPgSdTJgTJxI7leds cjLrDi5bgeJDjQDgqNgeUXLs8Ew4R3/JbEzj/nqJR7rO9tNvxY1G5g== =nIFO -----END PGP SIGNATURE-----
To unsubscribe, e-mail: trinity-users-unsubscribe@lists.pearsoncomputing.net For additional commands, e-mail: trinity-users-help@lists.pearsoncomputing.net Read list messages on the web archive: http://trinity-users.pearsoncomputing.net/ Please remember not to top-post: http://trinity.pearsoncomputing.net/mailing_lists/#top-posting
On Thursday 18 September 2014 23:40:31 Michael . wrote:
Lisi, the Debian vote, as far as I am aware, actually included all currently available options. In other words upstart and systemd were not the only choices available. The thing is 4 TC members (Canonical employees) voted for what they know, the other 4 voted for systemd (I personally think it was a deliberate vote against Canonical control) and that then left the chairmans vote which, as we know, went systemd (I also believe Garbee voted systemd to vote against Canonical control). I have read the discussion and I will try to find the actual vote, in order to see if my recollection is correct, and I will reply again if I am wrong to clarify things.
Thanks, Michael. That is very clear.
Lisi
-
Alexandre -
Curt Howland -
Dr. Nikolaus Klepp -
Lisi Reisz -
Michael . -
Mike Bird -
Steven D'Aprano -
Timothy Pearson