On 01/16/2015 04:25 PM, Richard Esplin wrote: > This forum post got me seriously considering the merits of systemd: > > https://bbs.archlinux.org/viewtopic.php?pid=1149530#p1149530 > > Though I am generally on the fence with regards to systemd, I think > there is a lot of value in increasing the consistency across Linux > distributions. If I'm going to have to deal > http://0pointer.net/blog/revisiting-how-we-put-together-linux-systems.htmlwith > systemd either way, I > hope it serves that purpose so that our effective user base is greater > and it is easier for 3rd party software to support my platform of choice.
Yes I can see how systemd could make it easier for 3rd party software to support all distros. Instead of writing a different init script for each distro, just ship a service file and away we go. But some argue that eliminates the diversity in distributions. In the end the only thing differentiating distros would be just a combination of package manager and desktop interface. Maybe that's not so bad. In the LAS interview, Lennart spent some time talking about other non-systemd -related things, such as weaknesses in how software is built, packaged, and shipped on Linux. He said the packaging systems we've used for years do work good for some core parts of a distro but they really are lousy for other things like applications. If software can be packaged and maintained by a distro life is pretty good. But in other cases it's not so good. For example in Ubuntu the owncloud package has not been maintained and should not be used. Projects that want to release binary distributions of their software have to either keep a build farm of different distros to build packages for, or make a tarball that bundles the specific libraries that they depend on such as Gtk. This is what firefox, LibreOffice, VirtualBox all do. And it kind of sucks, honestly. It is possible it install multiple versions of shared libraries in linux, but unfortunately that's not enough. Libraries rely on other data that's much harder to version (/usr/share/blah, etc). Poettering suggest perhaps the OS X framework model might be one we can learn from. There's a lot more to it as well that's very interesting. http://0pointer.net/blog/revisiting-how-we-put-together-linux-systems.html This is all pie in the sky still, but the end result would make different distros look more similar, from the app's point of view. Some would argue this is a horrible thing. But without it I just don't see the traditional software industry embracing Linux at all. Whereas they are embracing Android, and possibly to a lesser extend ChromeOS. Apparently the CoreOS uses ChromeOS as its base. And it doesn't have a package manager at all from what I understand. Anyway, I find it very interesting, and even exciting stuff, but I'm not here to rehash old arguments, so I will bow out now. /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
