Miles Fidelman <mfidel...@meetinghouse.net> writes: > I was watching the discussion on systemd fairly closely. I could be > wrong, but very little of the discussions over systemd seemed to reflect > folks who managed production servers, or kernel developers, or developers > of key backend software (Apache, MySQL, Postfix, Sympa, ...).
Well, for whatever it's worth, I was managing Debian production servers during the entire period of that debate (and for about ten years previous to that). I was the primary advocate for Stanford running its central infrastructure on Debian, so I'm familiar with the problems and arguments for and against using Debian in that sort of environment. Some of the other major voices in that debate manage far large production deployments than I did. My current employer uses Ubuntu in production, not Debian, for many of the typical reasons why people use Ubuntu over Debian, but from the perspective of systemd that's basically the same thing. Ubuntu went through essentially the same transition that we did. I do think distributions have some interesting challenges in the future, and what our users are asking from us is shifting. Containers and deep dependency programming ecosystems are both becoming more common, Go and Rust take a far different attitude towards how to assemble system components than C and C++ projects have historically, and cloud deployments are becoming far more common than hardware deployments for many of our users. One of the simultaneously fun and frustrating thing about this field is that problems are constantly shifting, and new ideas and new ways of doing things are constantly arising. Debian certainly will need to change and explore new and different corners of that, and feedback from day-to-day users of Debian both inside and outside the project will be very important to understand how to change. But, if anything, I think being *more* agile, not less, is where we can improve the most. And, of course, always trying to find ways in which we can have it all at once, where we can: provide a broad and inclusive platform for making a lot of different choices, so that we don't have to pick the best choices in advance and over-commit to one way of doing things. Which, among other things, calls for init system diversity, and I'm very glad to see that work continuing (although I personally still hope that someone will come up with a great init system that has the highly decoupled properties that people want but that isn't sysvinit with all of its well-known problems). I'll stop talking about this here since several folks are saying that we should keep init systems out of this conversation, and I'm not really helping. You just raised some points about the social impact of hard disagreements, and about how decision-making works in general in Debian, about which I have strong opinions and really wanted to reply. -- Russ Allbery (r...@debian.org) <http://www.eyrie.org/~eagle/>