On Lu, 03 mar 14, 14:29:16, Gian Uberto Lauri wrote: > Andrei POPESCU writes: > > Systemd can help a bit in making a little easier to have tools that > satisfy the (not so) basic need "to have this device mounted here if > it is plugged, otherwise go ahead with the bootstrap" for the > completely tech-unsavy user, but you can achieve this with system V > init. If you want to do it. This shouldn't be necessary. > > we (and by this I mean mostly debian-user subscribers) are tinkerers, > > otherwise we wouldn't be here. > > WARNING! If I got right your words, this is a damn narrow-sighted > point of view. If you stop a moment and and think, you may see how > many user of Debian GNU/Linux may exists that do not access this > mailing list. Or do not access any. Exactly what I was trying to say. systemd may not "really" be necessary for us, but what about... > > But, how am I going to do that for my > > father's laptop, which I *might* be able to access remotely? > > Excuse me, could you re-state this sentence. I am unable to understand > the point, sorry that is due my poor English skills. ... my father running Debian on his laptop? If he relocates to a foreign country, buys a 3G adapter and plugs it in what should happen? > > This is a joke right? If I tell a daemon to restart I want it > > restarted. Now. Anything else is like the tail wagging the dog. > > Sorry, no, or I could equally say "I want an Aston Martin parked in > the company yard. Now and anything else is like the tail wagging the > dog". Actually not. As I see it the computer is a tool built to do what *I* say and *when* I say it, not to create more work for me. Especially if it's only a Simple Matter of Programming (which turns out to not be so simple, since it took so many years to do it). > We agree that there are "some steps involved" between "wanting the > Astong Martin" and getting one. > > The same is for restarting a service (provided by a daemon). > > If you want a daemon to restart you either have a daemon that can > completely resets itself upon receiving, say, SIGHUP or you need to > terminate the previous instance and start a new one. > > And when you terminate a program you want to restart, you have to wait > for that program to be terminated to be sure all resources are > released. What if it doesn't do that and it just hangs? > And in this systemd has no more power than a script. It has to issue > the stopping signal, wait for the process to die and let free the > resources it used, and finally start a new one. What if it just won't die? systemd's answer to that is cgroups. Is this the "best" solution? I couldn't say. There's at least the problem that cgroups is Linux specific. AFAIU BSD jails could provide similar functionality, but chances of this ever being implemented in systemd are slim. Not so with OpenRC, which is much more modular and portable (which is why I think it *could* be the better solution in the long run). > And, AFAIK, if a process is not son of some other process then is son > of init by 'adoption'. But sysvinit is not actually used to manage processes, so this special power of PID 1 is wasted. > My opinion that the migration was too swift. I am almost sure that > a less "disruptive" way was possible provided some more time. > > systemd will be not as evil as I feared when I was first pointed to > some random document. It could add even some good in the long run. But > this too early migration will give some troubles to someone, I am in > this set. Debian is already late to the party. Just about every other major distribution/OS is already running something better than sysv-rc (and I'm including OpenRC in the "something better"). Kind regards, Andrei -- http://wiki.debian.org/FAQsFromDebianUser Offtopic discussions among Debian users and developers: http://lists.alioth.debian.org/mailman/listinfo/d-community-offtopic http://nuvreauspam.ro/gpg-transition.txt
Attachment:
signature.asc
Description: Digital signature