On Fri, Nov 14, 2014 at 10:01:39AM +0100, Josselin Mouette wrote: > The Wanderer <[email protected]> wrote: > I find that interesting. > > I recall being told that, while it is technically possible to compile > and use GNOME without systemd - specifically, without libpam-systemd > and > its backend infrastructure - doing so now loses so much functionality > that the result is barely (if at all) worth using. (This is a > paraphrase.) > > Is that not correct?
When logind was added as a dependency, it didn't depend on systemd. As release-team we mistakingly believed that it would stay independent. I suggested to Ian to look at the history while assuming people had the best intentions: - systemd: propose systemd as hard dependency - gnome: "no" - systemd: logind has really cool functionality fixing the stuff you dislike about ConsoleKit - gnome: "yes, let's get rid of ConsoleKit :-D" - kernel: cgroups needs to change - systemd: logind depends on systemd - gnome (some): "oh urgh" Now Lennart I know is very positive about systemd and wants it used everywhere. That's taken into account. Some stuff (timedated, hostnamed, etc) is really easy to implement independently. So I don't see any problem relying on that. Everyone knows he wants it everywhere. That's not what matters. When he wanted to have it everywhere, he got a no. Then he provided stuff that is useful, this resulted in optional parts being used. And IIRC, I think there were some warnings that logind was not independent (unlike timedated). But never really considered anything other than that it way a great replacement for ConsoleKit. > However, many pieces of GNOME depend directly on PolicyKit, which can > only use one of ConsoleKit or systemd-logind, the choice being made at > build time. > > In short: if you want to make GNOME in Debian work without systemd, you > need to make PolicyKit able to switch between both at runtime (like GDM > does). More important: Figure out technical details, be part of the technical discussions, respond technically why things should or should not be done. Be willing to do a bit of work if needed. Any change that might impact anything I usually announce, no matter if the change came from GNOME or not. See distributor-list archives. E.g. the tty switching which is now in logind and used for Wayland. Saying it should not be and leaving it at that (or e.g. voting for a GR) is not going to change things. Just join the discussion technically, propose a different technical solution. Be open to actually having to implement that. If someone offered an alternative possibility upstream then there is a possible different decision to be made. In some messages seem to suggest to leave upstream alone. That is just if you limit yourself to saying "do this/that". Another approach (meaning: technical description of alternative) I'd really appreciate. Ideally if the alternative seems feasible. -- Regards, Olav -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: https://lists.debian.org/[email protected]
