4 ms·
The scariest part no one is talking about is the feature creep. All software eventually dies; systemd will be no different. Some folks hope that happens immed
by VLM 13y ago
The scariest part no one is talking about is the feature creep. All software eventually dies; systemd will be no different. Some folks hope that happens immediately, some hope we're "stuck" with it for a couple generations, maybe 30 years. Regardless of exactly when, its going to be replaced someday, and massive feature creep seems intentionally designed to make the eventual changeover as painful as possible.
Someday in the future, either 2016 or 2066 someone is going to want to rearchitect the "thing that the kernel starts first" and is going to freak out at the insane feature creep... so you're telling me that I not only have to create something to start services which is no big deal, but also an entire logging system, and a hotplug handler, and a cgroup server, and ... WTF does this blasted thing have an MP3 player, a web server, and an IRC client in it too?
Its an awful design but probably survivable, at least temporarily, and once the flamewars die down a sensible replacement with a good design can be rolled out to replace it. But mid-flamewar with all manner of business / marketing money flying everywhere on all sides is not the ideal blue sky redevelopment time. As much as it pains me to say it, rolling out an init that has almost the worst possible inherent design possible does have some value to the ecosystem as a counterexample, here's what not to do with 2.0. Obviously its replacement will be better, I just hope it doesn't swing too far in the opposite direction as a reaction. The point of this paragraph is the most important thing to do at this time is to lay the seeds for the replacement, while not over-reacting to the disaster of systemd.
So, the replacement for systemd, to be rolled out sometime between 2015 and 2065, should ... (insert feature list here).
- marcosdumay 13y agoGreat, now what will think the unluky sould that'll need to replace Linux? The problem we are seeing is the framework vs. libraries dicotomy. Cut the features in too many packages, and you won't be able to use them anymore. Integrate them too much, and you won't be able to maintain them anymore. Anyway, just like the kernel, some day we must be able to say "launching the basic processes is a solved problem", and actualy solve it in a way that does not need to be replaced. We are not there, but we are in a point where the feature set is quite stable, so I guess we can improve usability a bit, at the expense of maintanability.