Autore: t.j.duchene Data: To: dng@lists.dyne.org, Joel Roth Oggetto: Re: [Dng]
No, the majority doesn't knows. Long life to the Scientific Method!
>You realize, of course, that Devuan developers are going to
>be up to their eyeballs to build up an infrastructure
>even *without* major changes to the underlying systems.
Of course, I do. I’m not suggesting anything radical here. As with my previous post, I am suggesting that this be evaluated at some point after Devuan is established and has made a good release. I firmly believe that it is the packaging methodology that will ensure that Devuan never ends up like Debian, where the core dev team makes decisions that cannot be undone. APT as it stands, and copying Debian simply does not guarantee that. I just think that discussing now, while we work toward a first release is appropriate so that we can agree that it needs future consideration.
>For perl, this is not particularly an issue, since perlbrew[1]
>allows the user to have multiple versions of
>perl installed in her $HOME directory.
Perl was just an example.
>APT is rather core to Debian, and unlikely, IMO to be
>rearchitected. AFAIK, source distros such as gentoo and CRUX
>do not allow this flexibility either.
I’m suggesting that perhaps it should. The only way to guarantee freedom of choice is to make it a core concept of the package management so that the developers never feel they have to make the system/sysv choice that Debian did. It’s not the first time that such a conversation has come up, and it certainly won’t be the last.
>I'm curious which package managers you think highly of.
I really don’t. I’ve seen a number of them, since I have used basically every major Linux out there, and APT definitely has room for improvement. It does a ton of things right, and a few things very, very badly. I know this because I’m packaged things before for different distributions.
>If Devuan takes off, there will likely be labor available
>to maintain initscripts, which don't usually change all that
>radically over time. My understanding is that systemd will replace
>init scripts with their own unique declarative files.
>No one will be re-writing init scripts for systemd.
I don’t expect there to be, but you really have to keep in mind how invasive system really is. We might be talking about just startup scripts at first with Jesse, but as time goes on we are going to have login, PAM, cron and other issues that will need to be addressed if Debian goes hip deep into systemd with each release.
Eventually, there will be no way to give users what they want and maintain package compatibility with Debian during the building process. No matter what happens, at some point in the future, I believe Devuan will become a completely separate derivative because it is clear enough that our goals and Debian’s are clearly dissimilar.