Will systemd make FreeBSD more popular?

Status
Not open for further replies.
I don't know if this is the right place to post it, but I found an interesting page: Has anybody ever heard of supervisord?
http://supervisord.org/
It seems that you don't need to replace init to have something like launchd in f.e. FreeBSD.
This program, and may be there are many others, allows you to load your process control deamon at startup.
I am no expert, but this would basically satisfy everyone, the ones who want to use FreeBSD as a tablet OS, care about power consumption, care about the remarks Hubbard made, and the ones who want to use it still as a server OS.

Same idea as runit, only runit has a smaller codebase and simpler setup.
 
Also, just for reference, init has nothing to do with power consumption, the practicality of FreeBSD as a tablet OS or anything else. That's just smoke/mirrors by Jordan Hubbard. He's blowing smoke because he wants FreeBSD to continue to be more like a little brother to OS X. If he and iXSystems get their way on this point, my fork will take form, and with it, I'll be disavowing any further code intrusion from Apple. Clang was a good, clean break, but that's not launchd or plists as a lingua franca, things that will fundamentally change the way FreeBSD works. About the only Mach-based OS I can stand to run is Tru64, and DEC was fantastic at keeping it true to the UNIX philosophy.
 
But that would then also mean that the GPL has (kind of) failed. That might be a little bit depressing for GPL advocates, to put it kindly.

<Raising my hand>

I used to be a GPL advocate. I honestly thought that the GPL did more to keep F/OSS from being ripped off by corporations, than BSD.

I will admit I was wrong. The GPL has failed.
 
<Raising my hand>

I used to be a GPL advocate. I honestly thought that the GPL did more to keep F/OSS from being ripped off by corporations, than BSD.

I will admit I was wrong. The GPL has failed.

<Also raises hand> Used to be a militant Linux, and then a militant Mac, then again a militant Linux user for many years. I was wrong. The GPL is inherently flawed, and OS X is as well.
 
Same idea as runit, only runit has a smaller codebase and simpler setup.

I'm a fan of runit myself, but I've only used it under Linux as a complete init system. It was always my understanding that the init implementation and process supervisor were separate, so using the latter doesn't require the former, but the online runit documentation isn't really clear on that. I might be overlooking something.
 
I'm a fan of runit myself, but I've only used it under Linux as a complete init system. It was always my understanding that the init implementation and process supervisor were separate, so using the latter doesn't require the former, but the online runit documentation isn't really clear on that. I might be overlooking something.

Runit's supervisor runs independent of PID 1. It includes its own init process if you want, but its not necessary to use it. You can easily use rc init with it.
 
Also, just for reference, init has nothing to do with power consumption, the practicality of FreeBSD as a tablet OS or anything else. That's just smoke/mirrors by Jordan Hubbard. He's blowing smoke because he wants FreeBSD to continue to be more like a little brother to OS X. If he and iXSystems get their way on this point, my fork will take form, and with it, I'll be disavowing any further code intrusion from Apple. Clang was a good, clean break, but that's not launchd or plists as a lingua franca, things that will fundamentally change the way FreeBSD works. About the only Mach-based OS I can stand to run is Tru64, and DEC was fantastic at keeping it true to the UNIX philosophy.

The thing with power consumption is my mistake not Hubbards (who by the the way gave a great talk). One could read out of my post that I suggest that power consumption has something to do with the init-software. Hubbard mentioned process control programs because unlike in servers, devices are plugged in and pulled out in phones and tablets and laptops all the time and the software has to respond to this. Start that or that demon if it is needed, kill it if it sits around doing nothing, start it again if it hanged itself, etc.. This, Hubbard said, has something to do with power consumption, which is obvious. You don't want to have 10 demons which are not used sucking on your DC power all the time. Only if a program asks for a particular demon, start it.
To be honest, I am a fan of power efficiency. I really would also like my desktop run as power efficient as possible. Therefore I enjoyed Hubbards talk so much. I feel bad when I forgot to turn off the light and come back in the evening and realize that the lamp burned the whole day wasting energy. And I feel not really good about running an OS which is not efficient, especially when the computer is turned on the whole day. Therefore I really am a fan of process control demons.
You may not like OSX, but it runs efficient, I tested this with my Macbook (mid 2011). Using OSX, you can write up to 7 hours- 7.5 hours a text read pdfs, look up this or that in the internet (not watching videos). I for testing purposes installed FreeBSD CURRENT on a second partition on it (amazingly it worked without any problems, I used rEFInd, the only problem is that there is no Broadcom driver :( ) and tested battery life without having installed X11, just idle. It said 4 hours and 30 minutes of battery life. That is a big difference.
I am absolutely sure that FreeBSD has the means to solve that problem, may be using runit and other programs from the mighty ports tree but a fresh install is not that power efficient. I want to be at least as good as 7-7.5 hours when I run my Macbook doing work. This is a very interesting challenge, even for people who run FreeBSD as a server. I want efficiently running servers :) because power=money and if I can get a more efficient software for running my server I choose that one :).
 
  • Thanks
Reactions: Oko
The thing with power consumption is my mistake not Hubbards (who by the the way gave a great talk). One could read out of my post that I suggest that power consumption has something to do with the init-software. Hubbard mentioned process control programs because unlike in servers, devices are plugged in and pulled out in phones and tablets and laptops all the time and the software has to respond to this. Start that or that demon if it is needed, kill it if it sits around doing nothing, start it again if it hanged itself, etc.. This, Hubbard said, has something to do with power consumption, which is obvious. You don't want to have 10 demons which are not used sucking on your DC power all the time. Only if a program asks for a particular demon, start it.
To be honest, I am a fan of power efficiency. I really would also like my desktop run as power efficient as possible. Therefore I enjoyed Hubbards talk so much. I feel bad when I forgot to turn off the light and come back in the evening and realize that the lamp burned the whole day wasting energy. And I feel not really good about running an OS which is not efficient, especially when the computer is turned on the whole day. Therefore I really am a fan of process control demons.
You may not like OSX, but it runs efficient, I tested this with my Macbook (mid 2011). Using OSX, you can write up to 7 hours- 7.5 hours a text read pdfs, look up this or that in the internet (not watching videos). I for testing purposes installed FreeBSD CURRENT on a second partition on it (amazingly it worked without any problems, I used rEFInd, the only problem is that there is no Broadcom driver :( ) and tested battery life without having installed X11, just idle. It said 4 hours and 30 minutes of battery life. That is a big difference.
I am absolutely sure that FreeBSD has the means to solve that problem, may be using runit and other programs from the mighty ports tree but a fresh install is not that power efficient. I want to be at least as good as 7-7.5 hours when I run my Macbook doing work. This is a very interesting challenge, even for people who run FreeBSD as a server. I want efficiently running servers :) because power=money and if I can get a more efficient software for running my server I choose that one :).

The approach Jordan Hubbard suggests would make us no more than an Apple's little brother OS.

The starting/stopping of daemons according to events can easily be done by a combination of daemons, see my below example:

User inserts USB drive. Kernel detects this and logs it, so devd creates a device file for it, and the automounter then automounts the drive. Similarly, if a user removes the drive without running umount, the kernel knows and logs it, so devd destroys the device file, which in turn forcibly unmounts the drive.

In launchd and systemd, this mostly takes place within their process groups. While this is simpler from an outside perspective, it inevitably leaves a small group of processes responsible for a large array of system critical functions. Myself, and many other users, see this as an affront to the UNIX philosophy, as well as a serious hazard to security and stability. A UNIX system in the traditional sense may have many points of failure, but its designed that if a few daemons or processes die that the system does not go down. The critical points of the system are the kernel and init, plus the network if you're on a server. By comparison, Windows has about 25 processes on a typical desktop that need to be running for the system to keep running interactively. If a single one of those go down, you have to reboot.

Fan control is not normally handled by the OS in well designed hardware, its handled by the firmware, and good firmware will equal efficiency. The fact is that exposing that much hardware to the OS isn't smart.

You also can't use your Macbook as a comparison here: BSD, Linux and Windows are unspecialised operating systems, meaning that they run on a wide range of hardware so they can't really afford to make hardware specific changes. OS X, by comparison, is the opposite. Its a specialised OS running on special hardware tailored to it. There's nothing wrong with this approach, see the SGI MIPS systems, but you can't make comparisons across both groups. Its literally APPLES to ORANGES.
 
Just found out, there was a recent itwire article on this subject:
I found quite a few dis-genuine statements in that interview starting with the sentence
At first glance it may even be difficult to tell them apart, we have almost all the same open-source software available
Kris is a smart guy so I will attribute those statements to the journalist. If the systemd becomes the King of the Linux land as it is looking now many of the commonly deployed software will stop working on BSDs. That question has been already addressed by OpenBSD in GSoC 2014 when the student was creating systemd like API so that programs like Gimp who are very close to the point of being systemd dependent can run in the future on OpenBSD.

Linux users might not have experience with BSDs but they are not stupid. It would be stupid for FreeBSD camp to claim vendor support (Kris talks about flash and Adobe reader in that interview) is such support doesn't exist. And indeed many people myself included have to run Linux due to the lack of vendor support (in my case MATLAB). UNIX desktop market is/was non-existing as correctly observed by Jordan Hubbard. There is some tiny corporate market for niche product like PC-BSD. On the another hand FreeBSD could easily take a serious bite from Linux on the server market if one of major players (Google, Microsoft, HP, IBM, Oracle) wishes to do so. However I am rather skeptical at the time when proprietary UNIX-es are either dead (IRIX), dying like Solaris or they are on the life support HP-Unix, AIX that anybody will put required investment into FreeBSD to make it viable competitor to Linux.
 
I'd say it's redhat that really pushed Linux in the server space, especially with their 10 year Dell partnership. I think if we landed collaborative support with any of the major server vendors (HP, Dell, Lenovo, etc) we'd see a boost. But as you stated we'd need huge vendor sustain that support. iXsystems could potentially be both (hardware/software, ie SUN) with TrueOS but they're focusing huge on Storage in 2015. I've looked at some of their development milestones, and its looking very good. FreeNAS 10 will be huge.
 
  • Thanks
Reactions: Oko
I'd say it's redhat that really pushed Linux in the server space, especially with their 10 year Dell partnership. I think if we landed collaborative support with any of the major server vendors (HP, Dell, Lenovo, etc) we'd see a boost. But as you stated we'd need huge vendor sustain that support. iXsystems could potentially be both (hardware/software, ie SUN) with TrueOS but they're focusing huge on Storage in 2015. I've looked at some of their development milestones, and its looking very good. FreeNAS 10 will be huge.
I think that people overestimate IXsystems. They are not major player in IT field even though they were in business since 1993 (essentially Berkeley folks who lost the job when Computer Systems Research Group was shut down founded the company). However you are raising very interesting point. FreeBSD should not try to compete with Linux across the board now as it is too much behind but in the particular market segment. You mentioned FreeNAS and storage appliances. That is a great example. FreeBSD is hands down superior OS for people who need to store their data. Linux people can continue to elect file system flavour of the day but I don't trust my data to any file system on Linux except perhaps old trusted SGI XFS file system.
 
However I am rather skeptical at the time when proprietary UNIX-es are either dead (IRIX), dying like Solaris or they are on the life support HP-Unix, AIX that anybody will put required investment into FreeBSD to make it viable competitor to Linux.

FWIW: I was doing some contract work for IBM last year. In part, I was helping with replacing their AIX systems with RHEL.

This was only at one, fairly small, installation. Although I got the idea that this is where IBM is heading. IBM seems to really like RHEL, it was on everybody's desktop (just like Lotus Notes).
 
Another, more recent article:

Sunday, February 15, 2015

A Prediction: 2020 the year of (PC-)BSD on the desktop

I am going to make a prediction right now that FreeBSD is going to take off in a big way on or before 2020, perhaps even to the point where it threatens Linux Desktop share.

http://lukewolf.blogspot.com/2015/02/a-prediction-2020-year-of-pc-bsd-on.html

Actually, that just looks like some blog post. FWIW: I came across this on slashdot. There is a fairly active discussion about this going on now.

http://bsd.slashdot.org/story/15/02/16/2355236/pc-bsd-set-for-serious-growth

I don't know if it means anything, but FreeBSD seems to be getting some attention in the pop-media.
 
I don't trust my data to any file system on Linux except perhaps old trusted SGI XFS file system.

I certainly don't anymore either after all the bizarre things that used to happen with my Linux PC NAS. Since I replaced the OS on it years ago, I've not had any of those issues again.

The blog article is made the same guy who was in here not too long ago nagging about why FreeBSD didn't have this and that luxury feature.
 
I still get queasy over my needing to make a decision on what to use for my new client's server OS. I decided I was going to be forced into choosing Linux over FreeBSD and then had to choose between Ubuntu and CentOS. Then other suggestions come flying in just making the choice harder. I KNOW FreeBSD is the better choice and only one basic reason is you can't pin Linux distributions down to any one solid thing. It's a constantly moving (meandering) target.

btw, yesterday I signed them up with a hosting company and I installed FreeBSD. I think I'll continue with that until someone who cares notices. I may get fired for not choosing IBM ... I mean Linux ... but I made the right choice.
 
So the day starts and the news tell me of new things happening. In this case, news about how to turn more users to *BSD systems.

It seems that the next step in the process for systemd to - ummm - enhance the user experience, is that they also want to integrate a boot loader. And containers. How did they manage to avoid that kitchen sink?
 
It still is a bit early for april fools day, is it? If that is real, there will be a shortage of popcorn.
 
April 1. will be when Linus praises the systemd-crowd and announces that he will resign the leading role in the kernel staff to these visionary and dynamic gentlemen. And he will have that text properly gendered, so it will be accepted by every person.
 
Status
Not open for further replies.
Back
Top