| Message ID | 20220221043034.2736010-1-francois.perrad@gadz.org |
|---|---|
| Headers | show |
| Series | portable services: another use case of BR | expand |
François, All, On 2022-02-21 05:30 +0100, Francois Perrad spake thusly: > This series allows to build a rootfs suitable for portable services with systemd. > > A new value of "Init system" is introduced : BR2_INIT_PORTABLE. > BR2_INIT_PORTABLE is like BR2_INIT_NONE, but with all FOO_INSTALL_INIT_SYSTEMD executed. > > A new type of skeleton is required : BR2_PACKAGE_SKELETON_INIT_PORTABLE. > > see https://systemd.io/PORTABLE_SERVICES/ I don't see anything that can't be already managed with existing configuration options. All that this new so-called init system does, is change some defaults to match a specific use-case. We had similar stuff back in the day (e.g. an option to define a router, one to define something else, and so on), and we removed all of them. Granted, portable services are slightly different, but still all that this series eventually does is change some defaults; it does not even enforce the settings so they are not even required. Even the new skeleton case can already be handled by providing a custom skeleton, and it is a trivial one, so even that is not needed. If you want to build a portable service, just build a system with systemd and the rest of the settings set as required for that use-case. So, my position is to reject this new pseudo-init system. Regards, Yann E. MORIN. > Francois Perrad (3): > package/skeleton-init-portable: new skeleton > portable: new init system > portable: tweak some default values > > DEVELOPERS | 1 + > fs/squashfs/Config.in | 1 + > package/Config.in | 1 + > package/ifupdown-scripts/Config.in | 2 +- > package/pkg-generic.mk | 2 ++ > package/skeleton-init-portable/Config.in | 7 ++++++ > .../skeleton-init-portable.mk | 23 +++++++++++++++++++ > system/Config.in | 18 +++++++++++---- > 8 files changed, 50 insertions(+), 5 deletions(-) > create mode 100644 package/skeleton-init-portable/Config.in > create mode 100644 package/skeleton-init-portable/skeleton-init-portable.mk > > -- > 2.32.0 > > _______________________________________________ > buildroot mailing list > buildroot@buildroot.org > https://lists.buildroot.org/mailman/listinfo/buildroot
François, all, On 2022-08-21 23:17 +0200, Yann E. MORIN spake thusly: > On 2022-02-21 05:30 +0100, Francois Perrad spake thusly: > > This series allows to build a rootfs suitable for portable services with systemd. > > > > A new value of "Init system" is introduced : BR2_INIT_PORTABLE. > > BR2_INIT_PORTABLE is like BR2_INIT_NONE, but with all FOO_INSTALL_INIT_SYSTEMD executed. > > > > A new type of skeleton is required : BR2_PACKAGE_SKELETON_INIT_PORTABLE. > > > > see https://systemd.io/PORTABLE_SERVICES/ > > I don't see anything that can't be already managed with existing > configuration options. Ah, I missed one thing: this allows installing systemd units bundled in Buildroot, without using systemd as a init system. This is flawed, IMO, because a lot of packages now install their own units as part of their install procedure, so without systemd enabled, those packages wold not install their units. So, in practice, to be really usefull, this would still need systemd to be enabled, and we're back to the situation where this series only brings a few defaults as policy. So, I'm still not convinced about it. Regards, Yann E. MORIN. > All that this new so-called init system does, is change some defaults to > match a specific use-case. We had similar stuff back in the day (e.g. an > option to define a router, one to define something else, and so on), and > we removed all of them. > > Granted, portable services are slightly different, but still all that > this series eventually does is change some defaults; it does not even > enforce the settings so they are not even required. > > Even the new skeleton case can already be handled by providing a custom > skeleton, and it is a trivial one, so even that is not needed. > > If you want to build a portable service, just build a system with > systemd and the rest of the settings set as required for that use-case. > > So, my position is to reject this new pseudo-init system. > > Regards, > Yann E. MORIN. > > > Francois Perrad (3): > > package/skeleton-init-portable: new skeleton > > portable: new init system > > portable: tweak some default values > > > > DEVELOPERS | 1 + > > fs/squashfs/Config.in | 1 + > > package/Config.in | 1 + > > package/ifupdown-scripts/Config.in | 2 +- > > package/pkg-generic.mk | 2 ++ > > package/skeleton-init-portable/Config.in | 7 ++++++ > > .../skeleton-init-portable.mk | 23 +++++++++++++++++++ > > system/Config.in | 18 +++++++++++---- > > 8 files changed, 50 insertions(+), 5 deletions(-) > > create mode 100644 package/skeleton-init-portable/Config.in > > create mode 100644 package/skeleton-init-portable/skeleton-init-portable.mk > > > > -- > > 2.32.0 > > > > _______________________________________________ > > buildroot mailing list > > buildroot@buildroot.org > > https://lists.buildroot.org/mailman/listinfo/buildroot > > -- > .-----------------.--------------------.------------------.--------------------. > | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | > | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | > | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | > | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | > '------------------------------^-------^------------------^--------------------' > _______________________________________________ > buildroot mailing list > buildroot@buildroot.org > https://lists.buildroot.org/mailman/listinfo/buildroot
On 22/08/2022 23:24, Yann E. MORIN wrote: > François, all, > > On 2022-08-21 23:17 +0200, Yann E. MORIN spake thusly: >> On 2022-02-21 05:30 +0100, Francois Perrad spake thusly: >>> This series allows to build a rootfs suitable for portable services with systemd. >>> >>> A new value of "Init system" is introduced : BR2_INIT_PORTABLE. >>> BR2_INIT_PORTABLE is like BR2_INIT_NONE, but with all FOO_INSTALL_INIT_SYSTEMD executed. >>> >>> A new type of skeleton is required : BR2_PACKAGE_SKELETON_INIT_PORTABLE. >>> >>> see https://systemd.io/PORTABLE_SERVICES/ >> >> I don't see anything that can't be already managed with existing >> configuration options. > > Ah, I missed one thing: this allows installing systemd units bundled in > Buildroot, without using systemd as a init system. In a portable service, /sbin/init is not used anyway, so the init system is irrelevant. The only real difference is that systemd and its dependencies are not installed to the target, which reduces size a little. > This is flawed, IMO, because a lot of packages now install their own > units as part of their install procedure, so without systemd enabled, > those packages wold not install their units. > > So, in practice, to be really usefull, this would still need systemd to > be enabled, and we're back to the situation where this series only > brings a few defaults as policy. Yeah, to be fully useful, we should probably replace all conditions on BR2_INIT_SYSTEMD (and many on BR2_PACKAGE_SYSTEMD) with something else... In addition, packages that actually require libsystemd (e.g. clamav) would need to select BR2_PACKAGE_SYSTEMD - and if you use any of those, the advantage is lost again. So, although the use case exists, I don't think the difference is important enough to bother, and this series addresses only a small part of the problem space. So I agree with Yann. Regards, Arnout > > So, I'm still not convinced about it. > > Regards, > Yann E. MORIN. > >> All that this new so-called init system does, is change some defaults to >> match a specific use-case. We had similar stuff back in the day (e.g. an >> option to define a router, one to define something else, and so on), and >> we removed all of them. >> >> Granted, portable services are slightly different, but still all that >> this series eventually does is change some defaults; it does not even >> enforce the settings so they are not even required. >> >> Even the new skeleton case can already be handled by providing a custom >> skeleton, and it is a trivial one, so even that is not needed. >> >> If you want to build a portable service, just build a system with >> systemd and the rest of the settings set as required for that use-case. >> >> So, my position is to reject this new pseudo-init system. >> >> Regards, >> Yann E. MORIN. >> >>> Francois Perrad (3): >>> package/skeleton-init-portable: new skeleton >>> portable: new init system >>> portable: tweak some default values >>> >>> DEVELOPERS | 1 + >>> fs/squashfs/Config.in | 1 + >>> package/Config.in | 1 + >>> package/ifupdown-scripts/Config.in | 2 +- >>> package/pkg-generic.mk | 2 ++ >>> package/skeleton-init-portable/Config.in | 7 ++++++ >>> .../skeleton-init-portable.mk | 23 +++++++++++++++++++ >>> system/Config.in | 18 +++++++++++---- >>> 8 files changed, 50 insertions(+), 5 deletions(-) >>> create mode 100644 package/skeleton-init-portable/Config.in >>> create mode 100644 package/skeleton-init-portable/skeleton-init-portable.mk >>> >>> -- >>> 2.32.0 >>> >>> _______________________________________________ >>> buildroot mailing list >>> buildroot@buildroot.org >>> https://lists.buildroot.org/mailman/listinfo/buildroot >> >> -- >> .-----------------.--------------------.------------------.--------------------. >> | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: | >> | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ | >> | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is no | >> | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v conspiracy. | >> '------------------------------^-------^------------------^--------------------' >> _______________________________________________ >> buildroot mailing list >> buildroot@buildroot.org >> https://lists.buildroot.org/mailman/listinfo/buildroot >
François, All, On 2022-02-21 05:30 +0100, Francois Perrad spake thusly: > This series allows to build a rootfs suitable for portable services with systemd. > > A new value of "Init system" is introduced : BR2_INIT_PORTABLE. > BR2_INIT_PORTABLE is like BR2_INIT_NONE, but with all FOO_INSTALL_INIT_SYSTEMD executed. > > A new type of skeleton is required : BR2_PACKAGE_SKELETON_INIT_PORTABLE. > > see https://systemd.io/PORTABLE_SERVICES/ Given the feedback on this series, and after discussing with the other maintainers, I've now marked this series as rejected. Sorry for that. If you believe the use-case is really important, we can reconsider it in the light of a newer and more complete series that addresses the issues that were raised in this review. Thank you! Regards, Yann E. MORIN. > Francois Perrad (3): > package/skeleton-init-portable: new skeleton > portable: new init system > portable: tweak some default values > > DEVELOPERS | 1 + > fs/squashfs/Config.in | 1 + > package/Config.in | 1 + > package/ifupdown-scripts/Config.in | 2 +- > package/pkg-generic.mk | 2 ++ > package/skeleton-init-portable/Config.in | 7 ++++++ > .../skeleton-init-portable.mk | 23 +++++++++++++++++++ > system/Config.in | 18 +++++++++++---- > 8 files changed, 50 insertions(+), 5 deletions(-) > create mode 100644 package/skeleton-init-portable/Config.in > create mode 100644 package/skeleton-init-portable/skeleton-init-portable.mk > > -- > 2.32.0 > > _______________________________________________ > buildroot mailing list > buildroot@buildroot.org > https://lists.buildroot.org/mailman/listinfo/buildroot
Le mar. 23 août 2022 à 18:46, Yann E. MORIN <yann.morin.1998@free.fr> a écrit : > François, All, > > On 2022-02-21 05:30 +0100, Francois Perrad spake thusly: > > This series allows to build a rootfs suitable for portable services with > systemd. > > > > A new value of "Init system" is introduced : BR2_INIT_PORTABLE. > > BR2_INIT_PORTABLE is like BR2_INIT_NONE, but with all > FOO_INSTALL_INIT_SYSTEMD executed. > > > > A new type of skeleton is required : BR2_PACKAGE_SKELETON_INIT_PORTABLE. > > > > see https://systemd.io/PORTABLE_SERVICES/ > > Given the feedback on this series, and after discussing with the other > maintainers, I've now marked this series as rejected. Sorry for that. > > Thanks for these feedbacks. François > If you believe the use-case is really important, we can reconsider it in > the light of a newer and more complete series that addresses the issues > that were raised in this review. > > Thank you! > > Regards, > Yann E. MORIN. > > > Francois Perrad (3): > > package/skeleton-init-portable: new skeleton > > portable: new init system > > portable: tweak some default values > > > > DEVELOPERS | 1 + > > fs/squashfs/Config.in | 1 + > > package/Config.in | 1 + > > package/ifupdown-scripts/Config.in | 2 +- > > package/pkg-generic.mk | 2 ++ > > package/skeleton-init-portable/Config.in | 7 ++++++ > > .../skeleton-init-portable.mk | 23 +++++++++++++++++++ > > system/Config.in | 18 +++++++++++---- > > 8 files changed, 50 insertions(+), 5 deletions(-) > > create mode 100644 package/skeleton-init-portable/Config.in > > create mode 100644 package/skeleton-init-portable/ > skeleton-init-portable.mk > > > > -- > > 2.32.0 > > > > _______________________________________________ > > buildroot mailing list > > buildroot@buildroot.org > > https://lists.buildroot.org/mailman/listinfo/buildroot > > -- > > .-----------------.--------------------.------------------.--------------------. > | Yann E. MORIN | Real-Time Embedded | /"\ ASCII RIBBON | Erics' > conspiracy: | > | +33 662 376 056 | Software Designer | \ / CAMPAIGN | ___ > | > | +33 561 099 427 `------------.-------: X AGAINST | \e/ There is > no | > | http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL | v > conspiracy. | > > '------------------------------^-------^------------------^--------------------' > _______________________________________________ > buildroot mailing list > buildroot@buildroot.org > https://lists.buildroot.org/mailman/listinfo/buildroot >