Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On Wed, Mar 26, 2014 at 11:38:40AM -0500, Kevin Toppins wrote: On 26 March 2014 10:13, Cameron Norman camerontnor...@gmail.com wrote: [...] That is pretty much impossible, according to the developers of the logind API and its single implementation. Perhaps a subset of the logind API for use by desktop environments / compositors would be more useful in this init and OS portability predicament. I think Matthias Clasen, a GNOME developer, actually recently expressed interest in this from a portable window manager and compositor's perspective. Ryan Lortie said he'd work with others on some minimal logind like API during 3.14 cycle. But focussed on entire GNOME stack, not just gnome-shell. Complication being GDM (=extra work). Implementation on non-Linux to be done by those developers (FreeBSD, etc; they seems to be ok with that) I can tell you right now, it is *vastly more difficult* to try to adapt programs modified to work with systemd in their current state, than it is to *revert* those programs to their pre-systemd state. You're so certain while so utterly wrong on so many levels it is pretty amusing and embarrassing at the same time. You said you don't easily get offended, but hopefully you do pickup some learnings here. -- Regards, Olav (GNOME release team) -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140328081951.ga17...@bkor.dhs.org
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
El Fri, 28 de Mar 2014 a las 1:19 AM, Olav Vitters o...@vitters.nl escribió: On Wed, Mar 26, 2014 at 11:38:40AM -0500, Kevin Toppins wrote: On 26 March 2014 10:13, Cameron Norman camerontnor...@gmail.com wrote: [...] That is pretty much impossible, according to the developers of the logind API and its single implementation. Perhaps a subset of the logind API for use by desktop environments / compositors would be more useful in this init and OS portability predicament. I think Matthias Clasen, a GNOME developer, actually recently expressed interest in this from a portable window manager and compositor's perspective. Ryan Lortie said he'd work with others on some minimal logind like API during 3.14 cycle. But focussed on entire GNOME stack, not just gnome-shell. Complication being GDM (=extra work). Implementation on non-Linux to be done by those developers (FreeBSD, etc; they seems to be ok with that) Thank you, that was who I was thinking of. Best regards, -- Cameron Norman
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 28 Mar 2014 03:40, Olav Vitters wrote: [...] I can tell you right now, it is *vastly more difficult* to try to adapt programs modified to work with systemd in their current state, than it is to *revert* those programs to their pre-systemd state. You're so certain while so utterly wrong on so many levels it is pretty amusing and embarrassing at the same time. You said you don't easily get offended, but hopefully you do pickup some learnings here. Did you understand that... There are (at least) two paths to take here, and this is referring to the more difficult of the two. The easier choice is the reversion then selective upgrade path. I did go on to explain why the first path was more difficult. Did you consider what I said there? Please, put why I'm so utterly wrong into writing here and let's see what you understand. -Kev -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/cadkoaxhnleetyo6-fgjqujaryhnq8nv6kaxu8bazyudvzvr...@mail.gmail.com
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On Fri, Mar 28, 2014 at 11:08:54AM -0500, Kevin Toppins wrote: On 28 Mar 2014 03:40, Olav Vitters wrote: [...] I can tell you right now, it is *vastly more difficult* to try to adapt programs modified to work with systemd in their current state, than it is to *revert* those programs to their pre-systemd state. You're so certain while so utterly wrong on so many levels it is pretty amusing and embarrassing at the same time. You said you don't easily get offended, but hopefully you do pickup some learnings here. Did you understand that... There are (at least) two paths to take here, and this is referring to the more difficult of the two. The easier choice is the reversion then selective upgrade path. Nope. I did go on to explain why the first path was more difficult. Did you consider what I said there? You claimed, you never explained. Ryan analysed, convinced maintainers, convinced other developers, convinced release team. There is a *huge* difference between the two. Please, put why I'm so utterly wrong into writing here and let's see what you understand. That is pointless; I don't care what you wanted to prove. Do the same amount of work others have done for starters, then we'll see. -- Regards, Olav -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140329020452.gb17...@bkor.dhs.org
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
Quoting Cameron Norman (camerontnor...@gmail.com): El Wed, 26 de Mar 2014 a las 9:03 PM, gustavo panizzo gfa g...@zumbi.com.ar escribió: On 03/26/2014 11:49 PM, Cameron Norman wrote: I wonder if dbus activation could be used to accomplish this. Of course, then one would not be able to put (in the case of Upstart) the socket bridge, dbus bridge, dbus, or anything those services need to boot into a cgroup, but one can still put stuff like Apache, lightdm/gdm/kdm/sddm, nginx, et al into a cgroup. Another option is to push the kernel maintainers to allow delegating parts of the cgroups tree to other processes, so that the init system could say you get a sub-hierarchy, you get a sub-hierarchy without the complication of multiple separate hierarchies. How do you suggest this integration with cgroups be done? i just want to put services inside cgroups, no socket activation, no dbus, no dbus activation. Haha. The problem there is that cgmanager uses dbus! So you need It doesn't use the system or session bus, though, just listens over its own unix socket. Dbus does not need to be started first. -serge -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140327142638.GC7658@sergelap
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 26 Mar 2014 12:30, Matthias Urlichs matth...@urlichs.de wrote: [...] But here is the vastly oversimplified technical argument... To the point of being neither technical nor valid. (Which admittedly was never in doubt even before I started reading.) What do you consider technical? Vastly oversimplified doesn't mean automatically wrong in this context. It means there are a huge amount more of valid technical points I can raise here, but under the requirements expressed, I had to fit it down to a page, and so I left out quite a lot. These arguments are still valid, even if they are but only a small tip of the iceberg. I think you will confirm that neither you, nor I, nor the guy who came up with the original idea actually understands how it works If understanding how systemd works is so much of a problem for you that you cannot even conceive of anybody, let alone its author, doing so then I'd like to suggest that debian-devel is not the right place for you. One of the problems of giving truncated information is that some people aren't aware of the steps one has already taken to establish the validity of the argument. Lennart Poettering doesn't really understand what systemd is, and as a consequence, how it works. I tested this out myself. You can read up on how I determined this if you want, but it is a truth in its own right. I'd suggest an alternate mailing list, but I'm afraid I'd offent both you and the other participants of that list, should I do so. You do not have to fear that you will offend me. But, there are specific reasons why debian-devel is the primary target here. Again... information I left out in the condensed version. -Kev -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CADkoAxj+rFQXvGhv-BHF5pX0DZwMV=gyioaf_d8j8be-ywz...@mail.gmail.com
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 03/25/2014 12:42 AM, Kevin Toppins wrote: Sorry for the intrusion into your world, but this *needed* to be said, and needed to be said on *this* specific list. Not correct. We didn't need another iteration of such a post. - the *future* of linux actually *does* depend on what - you - *do* here Correct. Which doesn't include reading or writing such a message. If you want thing to move on, stop posting useless messages, and start working on alternatives. For example, helping adding more features to OpenRC would certainly help a way more than this post. Writing an independent, init system agnostic, logind API compatible daemon would be another good thing to do. Thomas Goirand (zigo) -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/5332ae91.9070...@debian.org
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On Wed, Mar 26, 2014 at 3:40 AM, Thomas Goirand z...@debian.org wrote: On 03/25/2014 12:42 AM, Kevin Toppins wrote: Writing an independent, init system agnostic, logind API compatible daemon would be another good thing to do. That is pretty much impossible, according to the developers of the logind API and its single implementation. Perhaps a subset of the logind API for use by desktop environments / compositors would be more useful in this init and OS portability predicament. I think Matthias Clasen, a GNOME developer, actually recently expressed interest in this from a portable window manager and compositor's perspective.
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 26 March 2014 05:40, Thomas Goirand z...@debian.org wrote: [...] If you want thing to move on, stop posting useless messages, and start working on alternatives. For example, helping adding more features to OpenRC would certainly help a way more than this post. I am going to have to respectfully disagree with you on my post being useless. First off, let's realize that we have more than one problem here. The actual implementation work that you indicate should be done is a valid point. We are both in agreement there. However, there exists an even bigger problem to be tackled. You can come up with all the solutions you want, but until it is *widely understood* that your solutions are *needed*, people tend to ignore and dismiss you. You can clearly see that happening in the responses I got back in Nov 2012. You first have to help people *understand* the problem and given how all the other topics on systemd being a failure *still didn't stop* debian's progress with using it I decided a very different perspective needed to be introduced. And I had to wait a while for things to get bad enough for people to see some validity in what I am saying. And while *you* might understand systemd is a problem, it is *objectively evident* that most do not, given the recent momentum to further adopt systemd by the linux community at large. My post is an analysis of systemd from an engineer's point of view. And systemd *violates* every engineering principle I spent years in college to learn. The biggest problem is awareness and that is the primary purpose of my post. And having it discussed in closed or private circles does not help mass awareness. It needs to be out in the open where everyone can read it. But it will have the most traction here. Hence why debian-devel was the primary target all along. -Kev -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CADkoAxhLo5fmJHEukhct2WB1ATH6a4yf_k=kjtm9qvc2gnk...@mail.gmail.com
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 26 March 2014 10:13, Cameron Norman camerontnor...@gmail.com wrote: [...] That is pretty much impossible, according to the developers of the logind API and its single implementation. Perhaps a subset of the logind API for use by desktop environments / compositors would be more useful in this init and OS portability predicament. I think Matthias Clasen, a GNOME developer, actually recently expressed interest in this from a portable window manager and compositor's perspective. I can tell you right now, it is *vastly more difficult* to try to adapt programs modified to work with systemd in their current state, than it is to *revert* those programs to their pre-systemd state. There is a huge amount of unknown design parameters that would need to be known to adapt those programs effectively, and since systemd clearly lacks any coherent design, you will not be able to identify coherent ways to fix this. That is your signal to abandon that path right there. Trust me on this, you don't want to try to patch the *current* versions. And what about all the non-systemd improvements made to those programs over the years they worked with systemd? For that, it is *orders of magnitude easier* to take the improvements and adapt them to the pre-systemd version of the program. This is why I specifically mentioned that Debian perform a pre-systemd reversion to escape this mess. Once you have reverted, you can then figure out how to apply the *useful* upgrades to the old programs. This is the easiest path you can take. -Kev -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CADkoAxh2_GyySKzTVOmdTvzSMy8w=pjkhu1u4vj76jtvw-y...@mail.gmail.com
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 26/03/14 17:13, Kevin Toppins wrote: I am going to have to respectfully disagree with you on my post being useless. With the hope of contributing constructive criticism, I'll answer that. As far as the systemd vs. upstart discussion, I was leaning in upstart (more precisely, against systemd). As such, your email was very interesting to me. Unfortunately, it was unreadable. You said you'll start with background, but instead of providing technical background, you provided meaningless and irrelevant he said, I said arguments. Trying to skim ahead to find where the meat starts did not easily detect such a point. At this point, I simply assumed the email had nothing more to say. If I'm wrong, feel free to answer with the technical gist of your arguments. If you want me to read it, please adhere to the following guidelines: * No more than one page. * No *asterisks* and - arrows. * No references to previous discussions, or other people's arguments for/against systemd. I believe in free discussion. As such, feel free to ignore these suggestions, just as I'll feel free to ignore your email if you do so. Shachar
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 26 March 2014 13:42, Shachar Shemesh shac...@debian.org wrote: [...] As far as the systemd vs. upstart discussion, I was leaning in upstart (more precisely, against systemd). As such, your email was very interesting to me. Unfortunately, it was unreadable. You said you'll start with background, but instead of providing technical background, you provided meaningless and irrelevant he said, I said arguments. Trying to skim ahead to find where the meat starts did not easily detect such a point. At this point, I simply assumed the email had nothing more to say. If I'm wrong, feel free to answer with the technical gist of your arguments. If you want me to read it, please adhere to the following guidelines: No more than one page. No *asterisks* and - arrows. No references to previous discussions, or other people's arguments for/against systemd. First, here is a version with the asterisks removed. It might be visually easier to read. https://lists.debian.org/debian-devel/2014/03/msg00449.html Second, some concepts need a lot of information communicated to make sense to those who are not already familiar with the concept. We don't send people to college for a day and expect them to grasp 4 years of higher education. There are some limits on Human learning that you have to respect. But here is the vastly oversimplified technical argument... The test of comprehension is... if you cannot put an idea into writing, then you do not understand that idea well enough to be of any practical use. If that idea is a program... this means you do not actually have control of the program when implemented. Our ability to control things is directly dependent on our knowledge of how that thing operates. With knowledge, comes the ability to manipulate the thing to suite our purposes. Please, tell me what systemd is, fitting its entire functionality as expressed as one single concept. That does not mean it has to be one sentence, but it does mean you cannot group different concept together and simply give that as an answer. Grouping them together is saying what it does, not what it is. Big difference. I think you will confirm that neither you, nor I, nor the guy who came up with the original idea actually understands how it works, and we will not actually have control of it in situations where we need to control it. And so we need to pull it out of linux. -Kev -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/CADkoAxgmfwpKw-cx4abjQZXK0d+-VgKzbu1Rugk=xtp3efk...@mail.gmail.com
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
Excerpts from Kevin Toppins's message of 2014-03-26 13:00:22 -0700: On 26 March 2014 13:42, Shachar Shemesh shac...@debian.org wrote: [...] As far as the systemd vs. upstart discussion, I was leaning in upstart (more precisely, against systemd). As such, your email was very interesting to me. Unfortunately, it was unreadable. You said you'll start with background, but instead of providing technical background, you provided meaningless and irrelevant he said, I said arguments. Trying to skim ahead to find where the meat starts did not easily detect such a point. At this point, I simply assumed the email had nothing more to say. If I'm wrong, feel free to answer with the technical gist of your arguments. If you want me to read it, please adhere to the following guidelines: No more than one page. No *asterisks* and - arrows. No references to previous discussions, or other people's arguments for/against systemd. First, here is a version with the asterisks removed. It might be visually easier to read. https://lists.debian.org/debian-devel/2014/03/msg00449.html Second, some concepts need a lot of information communicated to make sense to those who are not already familiar with the concept. We don't send people to college for a day and expect them to grasp 4 years of higher education. There are some limits on Human learning that you have to respect. But here is the vastly oversimplified technical argument... The test of comprehension is... if you cannot put an idea into writing, then you do not understand that idea well enough to be of any practical use. If that idea is a program... this means you do not actually have control of the program when implemented. Our ability to control things is directly dependent on our knowledge of how that thing operates. With knowledge, comes the ability to manipulate the thing to suite our purposes. Please, tell me what systemd is, fitting its entire functionality as expressed as one single concept. That does not mean it has to be one sentence, but it does mean you cannot group different concept together and simply give that as an answer. Grouping them together is saying what it does, not what it is. Big difference. Kevin, it would be quite helpful for those who accept this challenge if you could do the same for the pieces of the stack that systemd is meant to replace or is (was?) competing with. So - Sysvinit is ... - Upstart is ... - OpenRC is ... Or since your argument is that Linux fits into this: Linux is ... Thanks. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/1395865553-sup-7...@fewbar.com
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
Hi, Kevin Toppins: But here is the vastly oversimplified technical argument... To the point of being neither technical nor valid. (Which admittedly was never in doubt even before I started reading.) I think you will confirm that neither you, nor I, nor the guy who came up with the original idea actually understands how it works If understanding how systemd works is so much of a problem for you that you cannot even conceive of anybody, let alone its author, doing so … then I'd like to suggest that debian-devel is not the right place for you. I'd suggest an alternate mailing list, but I'm afraid I'd offent both you and the other participants of that list, should I do so. -- -- Matthias Urlichs signature.asc Description: Digital signature
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 03/26/2014 07:40 AM, Thomas Goirand wrote: If you want thing to move on, stop posting useless messages, and start working on alternatives. For example, helping adding more features to OpenRC would certainly help a way more than this post. going offtopic here, do you know if there is any plan to use cgmanager with openrc, i really like the idea of putting each service in it's own cgroup -- 1AE0 322E B8F7 4717 BDEA BF1D 44BB 1BA7 9F6C 6333 -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/533387ec.2090...@zumbi.com.ar
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
El Wed, 26 de Mar 2014 a las 7:07 PM, gustavo panizzo gfa g...@zumbi.com.ar escribió: On 03/26/2014 07:40 AM, Thomas Goirand wrote: If you want thing to move on, stop posting useless messages, and start working on alternatives. For example, helping adding more features to OpenRC would certainly help a way more than this post. going offtopic here, do you know if there is any plan to use cgmanager with openrc, i really like the idea of putting each service in it's own cgroup I was thinking about how to do something like this without requiring cgmanager to be started before the init system or moving the cgroups management into the init system itself. I wonder if dbus activation could be used to accomplish this. Of course, then one would not be able to put (in the case of Upstart) the socket bridge, dbus bridge, dbus, or anything those services need to boot into a cgroup, but one can still put stuff like Apache, lightdm/gdm/kdm/sddm, nginx, et al into a cgroup. Another option is to push the kernel maintainers to allow delegating parts of the cgroups tree to other processes, so that the init system could say you get a sub-hierarchy, you get a sub-hierarchy without the complication of multiple separate hierarchies. How do you suggest this integration with cgroups be done? Best regards, -- Cameron Norman
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
On 03/26/2014 11:49 PM, Cameron Norman wrote: El Wed, 26 de Mar 2014 a las 7:07 PM, gustavo panizzo gfa g...@zumbi.com.ar escribió: On 03/26/2014 07:40 AM, Thomas Goirand wrote: If you want thing to move on, stop posting useless messages, and start working on alternatives. For example, helping adding more features to OpenRC would certainly help a way more than this post. going offtopic here, do you know if there is any plan to use cgmanager with openrc, i really like the idea of putting each service in it's own cgroup I was thinking about how to do something like this without requiring cgmanager to be started before the init system or moving the cgroups management into the init system itself. i don't see any problem starting cgmanager after init, i don't see much value on a big init or an init daemon confined by a cgroup. I wonder if dbus activation could be used to accomplish this. Of course, then one would not be able to put (in the case of Upstart) the socket bridge, dbus bridge, dbus, or anything those services need to boot into a cgroup, but one can still put stuff like Apache, lightdm/gdm/kdm/sddm, nginx, et al into a cgroup. Another option is to push the kernel maintainers to allow delegating parts of the cgroups tree to other processes, so that the init system could say you get a sub-hierarchy, you get a sub-hierarchy without the complication of multiple separate hierarchies. How do you suggest this integration with cgroups be done? i just want to put services inside cgroups, no socket activation, no dbus, no dbus activation. i would use it for servers, apache and friends, what i would really like is to be able to run multiple instances of the same service each on it's own cgroup. something like su - user -g cgroup_name -c command or a flag to start-stop-daemon, cgroups could be created in advance by an init script (a Required-Start in lsb slang) just my 0.02$ of what i would use as sysadmin Best regards, -- Cameron Norman -- 1AE0 322E B8F7 4717 BDEA BF1D 44BB 1BA7 9F6C 6333 -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/5333a2fe.2040...@zumbi.com.ar
Re: stop posting useless cruft and get to work (systemd and Linux are *fundamentally incompatible* - and I can prove it)
El Wed, 26 de Mar 2014 a las 9:03 PM, gustavo panizzo gfa g...@zumbi.com.ar escribió: On 03/26/2014 11:49 PM, Cameron Norman wrote: I wonder if dbus activation could be used to accomplish this. Of course, then one would not be able to put (in the case of Upstart) the socket bridge, dbus bridge, dbus, or anything those services need to boot into a cgroup, but one can still put stuff like Apache, lightdm/gdm/kdm/sddm, nginx, et al into a cgroup. Another option is to push the kernel maintainers to allow delegating parts of the cgroups tree to other processes, so that the init system could say you get a sub-hierarchy, you get a sub-hierarchy without the complication of multiple separate hierarchies. How do you suggest this integration with cgroups be done? i just want to put services inside cgroups, no socket activation, no dbus, no dbus activation. Haha. The problem there is that cgmanager uses dbus! So you need dbus installed + started before you can use cgroups with later kernels. i would use it for servers, apache and friends, what i would really like is to be able to run multiple instances of the same service each on it's own cgroup. something like su - user -g cgroup_name -c command or a flag to start-stop-daemon, cgroups could be created in advance by an init script (a Required-Start in lsb slang) I think cgexec is what you are looking for here. This would break with later versions of the linux kernel, though, because you need one centralized writer (e.g. cgmanager or systemd). -- Cameron Norman