Is systemd unix? [LWN.net]

Is systemd unix?

Posted Apr 26, 2012 5:07 UTC (Thu) by ebiederm (subscriber, #35028)
Parent article: The return of the Unix wars?

Only looking at Ubuntu from a technical perspective is short sighted.

We have had a lot of strange developments over the years. Udev has gone from being the great tool of userspace device name policy to the daemon that oly changes the permissions on devfs device nodes. What???? That isn't what we agreed to.

When looked at from a certain light the choices of the systemd developers are not at all in the unix tradition, and there seems to be disdain for the unix tradition of doing one thing and doing it well. Furthermore systemd seems to be rewritting so many things so fast that we are losing hard won lessons from our history.

Will RHEL7 based on systemd and gnome3.x and Wayland be unix? Will it run my old unix applications? We have seen enough disdain for backwards compatibility that I am in serious doubt.

On the flip side there are a lot of stories of running a normal arm distribution in a chroot on arm phones that it seems clear there is still one kernel pulling these things together.

Every distribution has a libc.

So I expect we will continue to have the ability to run each others binaries if we care and open source is still an ideal so the differences do not look like the propritary lock in of the unix wars but more like the divrsity that leads to new and better solutions that we can still pick a best of breed out of. So I don't see this as a return to the unix wars but as something new and different in this world.

I do worry about the scrapping of evolutionary change, and instead this great jump
into revolutionary change that has left me unable to find a non-buggy desktop except debian stable. Last I looked our latest and greatest things suck. Perhaps somewhere in this grat diversity someone will get it right.

(Log in to post comments)

Is systemd unix?

Posted Apr 26, 2012 7:41 UTC (Thu) by ersi (subscriber, #64521) [Link]

Well, I'd say you're both right.

It's all about the frame of mind when rewriting whatever code we're referencing to.

If you're rewriting to just throw out the past, you're in my opinion likely to not care what the current code does in detail. Which likely ends up with "new" history lessons coming - when you need to iterate over this rewrite and fix stuff you didn't think about.

If you're rewriting a section/whole program to merge/extend it, I think you're more likely to try to get a good grasp of what it does to a larger extent and keep bug fixes and quirks while baking new code and refactoring some old code.

Is systemd unix?

Posted Apr 27, 2012 22:24 UTC (Fri) by yoe (subscriber, #25743) [Link]

Actually that's not true, for many of the reasons explained here:

http://www.joelonsoftware.com/articles/fog0000000069.html

Is systemd unix?

Posted Apr 26, 2012 8:22 UTC (Thu) by ebirdie (subscriber, #512) [Link]

The above comments are excellent additions (in lack of a better word to addition) forcing me to comment.

tajyrink brings me to a question in the importance of brands, why there isn't yet a brand for Linux/FOSS technologies to apply for by manufacturers? The brand could use the ideas and message, what Microsoft has used for ages: "based on NT technology", "based on Microsoft technologies". I don't remember exact wordings from Microsoft and can't go after them right now, but similar approach/message with an example like "build on tried and true FOSS tech with Open Source ideals" could be useful.

ebiederm expands the view excellently above. There must be hundreds examples fitting to conversation and I'd like to bring my latest concerns from Open Source/Linux storage technologies. About five years ago I was wondering and asking the same questions, what Russell Coker asked in his blog this year: Reliability of RAID. Russell has concluded that ZFS and BTRFS are the way of future in storage. I agree on his concerns, how to be sure that data sitting on platter stays valid and how to detect, when the underlaying piece of technology starts to melt down. But I don't agree on his conclusions that stuffing everything to file systems is the right thing to do. I like the Unix tool approach here as well and that is what I see in current crop of storage technologies in device monitoring with technologies like smartmon / LVM / MD software RAID / file systems / some application level data integrity tools for data integrity / object storage solutions / network storage solutions. That is quite a bunch of technologies to play with and they don't play THAT well together eg., what I can tell from experience, slicing and dicing file systems ie. resizing them with LVM tools has not worked for me out of the box. I think it is not to blame LVM developers, but a sign of problem in Open Source technology community working as a whole producing a coherent working storage stack by the Unix tool approach. In sense of unix tool concept I find the storage stack lacking pipe, stdin/-out, regular expressions and argument syntax/semantics enabling the stack glued together by seemingly independent projects like LVM or MD. In this respect I see ZFS and BTRFS very appealing practical solutions. I hope my readers can see my point in looking at the whole and not derail from details.

In my wondering about RAID and its purpose I concluded that it is not RAID's job to care about data integrity but to provide continuance and uninterruptibility in case of maintenance or disaster. Some other tools must do the job detecting signs of rising problems, identifying the problem to a disk or a controller and either issuing hot-replace event to RAID subsystem or a warning to system logs for admin to spot for action.

In the above I must place my thanks to Neil Brown's articles both here at LWN.net and in his blog.

Is systemd unix?

Posted Apr 26, 2012 14:52 UTC (Thu) by drag (subscriber, #31333) [Link]

> Will RHEL7 based on systemd and gnome3.x and Wayland be unix? Will it run my old unix applications? We have seen enough disdain for backwards compatibility that I am in serious doubt.

In what way does Wayland or SystemD break legacy applications?

I see a lot of people throwing around terms like 'Unix way' and things like that, but I don't see it. I don't see it at all. Same thing people talk about the portability of using init scripts and such. I don't see that at all either.

The scripts and commands that I use to managing the start up of processes is different from Redhat to Debian. They are also different from what FreeBSD and OpenBSD use and they are different then what I see in Solaris. All of it it is very different.

SystemD still supports legacy scripts just as well as anything else ever did. Wayland will happily run all your X applications. etc etc.

I mean when Perl came out did people complain that it wasn't 'Unix'. That it wasn't 'do one thing and do it well' and that it was a conglomeration of different approaches in a generic multipurpose language and abandoned much of the pipe-passing and shell utilities and all that?

Is systemd unix?

Posted Apr 26, 2012 18:03 UTC (Thu) by rfunk (subscriber, #4054) [Link]

"I see a lot of people throwing around terms like 'Unix way' and things like that, but I don't see it. I don't see it at all."

The complaint is that systemd doesn't just take over for the old init system. It's also taking over for the old inetd, udev, syslog, and probably others. The "Unix" way is not to try to do too much in one subsystem or process, so this consolidation makes people nervous.

"The scripts and commands that I use to managing the start up of processes is different from Redhat to Debian. They are also different from what FreeBSD and OpenBSD use and they are different then what I see in Solaris. All of it it is very different."

I've run all of those systems, plus Ubuntu and some others (both Linux and commercial Unix) better forgotten, in the last 15+ years, and all but the BSDs have let me do "/etc/init.d/foo start" and "/etc/init.d/foo stop" to control daemons. (Since the advent of upstart, Ubuntu complains about it, but still does it.) And all but the BSDs have allowed basically the same procedure for making a daemon start on bootup (which minor variations in directory structure and runlevel definitions). All of it is very similar.

"I mean when Perl came out did people complain that it wasn't 'Unix'."

Actually people did, a bit. Ever heard the term "Swiss army chainsaw"? But they still used it. Eventually Perl embraced modularity, however, as have its many successors.

From my perspective, it seems like systemd is being pushed before it's widely considered ready to take over such fundamental tasks. Wayland is not yet being pushed, and seems to be trying to address the concerns first. Of course, Wayland has an advantage of trying to replace something that was already trying to do too much, and possibly even reducing the scope slightly.

Is systemd unix?

Posted Apr 26, 2012 18:25 UTC (Thu) by dlang (✭ supporter ✭, #313) [Link]

This is made worse by the fact that the replcements don't actually do everything that was possible (and done in practice) before, they only implement that parts that the new authors consider 'important'

sometimes they are 'considerate' enough to allow us to continue to run the old software, but only as a second class citizen doing whatever the new software allows it to do.

Is systemd unix?

Posted Apr 26, 2012 18:45 UTC (Thu) by jspaleta (subscriber, #50639) [Link]

wow... this conversation has gone too abstract to be useful even as constructive criticism. I'm not even sure what particular pieces of software or functionality you are complaining about.

Oh and...get off my lawn.

-jef