4 ms·

It appears you may be new to this whole "Unix" thing. What made unices popular was their ubiquitousness. If you had a BSD machine on the same network as a Linux

It appears you may be new to this whole "Unix" thing. What made unices popular was their ubiquitousness. If you had a BSD machine on the same network as a Linux or SunOS machine, they can communicate using standard protocols. Further, the applications themselves were portable between POSIX implementations. The Unix Way is to have small programs that do one thing extremely well; when an application (with the exception of emacs) starts emulating Zawinski's Law, they're usually broken up into smaller pieces. I do not want one app that handles cd, ls, pwd, rm, mv, etc. I suspect that if Poettering/Sievers released an app (let's call it "fileD") that did exactly that, the Unix community would gather the torches and pitchforks to go put an end to the nonsense. Kinda like they're doing right now with systemd. A Unix that does not implement or respect POSIX is not Unix. And if the major GNU/Linux developers wanted to try a different direction, they can. But they don't. The majority of developers want to keep Linux this way. two engineers from Red Hat want it this way, and politics are driving it's adoption rather than technical merit.

cbd1984 12y ago
> Further, the applications themselves were portable between POSIX implementations. It appears you may be new to this whole "Unix" thing. Application source portability was never absolute, not by a long shot, back when Sun and SGI and HP were still relevant to what Unix was. They all had value-add APIs, little different ways of doing things which were deliberately different from everyone else's to encourage lock-in. And even disregarding that, the BSD/SysV split was even stronger then than it is now, with heterodox APIs like STREAMS still in use. Linux and the Open Source BSDs are still different, but the standards are better-followed and the auto-configuration tools are better.
- anonbanker 12y ago
  > It appears you may be new to this whole "Unix" thing. I understand that you're doing this out of malice, and are using "reflection" in order to make yourself feel superior, and to take me down a peg, but I do regret using that sentence in the GP. The rest of the post was completely accurate, and that sentence diminished it. However, I'll have to disagree with the rest of your reply; The BSD/SysV split, while large, is fairly comparable to what's going on today with systemd. Poettering has made his contempt for *BSD very clear multiple times, and is actively trying to extinguish compatibility with it. GNOME needs massive work to play nice with any BSD now, thanks to Lennart. Porting between SunOS/IRIX/HP-UX was far less work than GNOME now requires.
digi_owl 12y ago
One binary or multiple binaries is a distraction. The issue is how they talk to each other. There already exist a single binary of what you describe, Busybox. The interesting thing is that you could probably mix busybox and the individual coreutils and nothing would break. This because the way they talk to each other, in steams of text via standard out and standard in, is defined and fixed. They also use those very same interfaces to talk to third parties.
- anonbanker 12y ago
  Thank you for this reply. I would probably argue that Busybox isn't really the unix way, but I'm tempted to adopt your viewpoint of fixed defined protocols to talk internally and externally as my stance going forward. Sadly, systemd still fails in this regard, whereas Busybox is a much better citizen.

cbd1984 12y ago

> Further, the applications themselves were portable between POSIX implementations. It appears you may be new to this whole "Unix" thing. Application source portability was never absolute, not by a long shot, back when Sun and SGI and HP were still relevant to what Unix was. They all had value-add APIs, little different ways of doing things which were deliberately different from everyone else's to encourage lock-in. And even disregarding that, the BSD/SysV split was even stronger then than it is now, with heterodox APIs like STREAMS still in use. Linux and the Open Source BSDs are still different, but the standards are better-followed and the auto-configuration tools are better.

anonbanker 12y ago
> It appears you may be new to this whole "Unix" thing. I understand that you're doing this out of malice, and are using "reflection" in order to make yourself feel superior, and to take me down a peg, but I do regret using that sentence in the GP. The rest of the post was completely accurate, and that sentence diminished it. However, I'll have to disagree with the rest of your reply; The BSD/SysV split, while large, is fairly comparable to what's going on today with systemd. Poettering has made his contempt for *BSD very clear multiple times, and is actively trying to extinguish compatibility with it. GNOME needs massive work to play nice with any BSD now, thanks to Lennart. Porting between SunOS/IRIX/HP-UX was far less work than GNOME now requires.

digi_owl 12y ago

One binary or multiple binaries is a distraction. The issue is how they talk to each other. There already exist a single binary of what you describe, Busybox. The interesting thing is that you could probably mix busybox and the individual coreutils and nothing would break. This because the way they talk to each other, in steams of text via standard out and standard in, is defined and fixed. They also use those very same interfaces to talk to third parties.

anonbanker 12y ago
Thank you for this reply. I would probably argue that Busybox isn't really the unix way, but I'm tempted to adopt your viewpoint of fixed defined protocols to talk internally and externally as my stance going forward. Sadly, systemd still fails in this regard, whereas Busybox is a much better citizen.