curious... Linux vs BSD ?
Just curious about state of Linux vs BSD... I use Linux, but examining my use case carefully, I don't really "use" Linux. I use applications and tools found in Linux distro. These applications and tools can also be found in BSD distro. To those who knows/uses both BSD and Linux... Should I learn BSD, and which one? -- William
On Wed, Sep 28, 2016 at 11:24:13PM -0400, William Park via talk wrote:
Just curious about state of Linux vs BSD... I use Linux, but examining my use case carefully, I don't really "use" Linux. I use applications and tools found in Linux distro. These applications and tools can also be found in BSD distro.
To those who knows/uses both BSD and Linux... Should I learn BSD, and which one?
Well my personal opinion is that working with BSD is too painful (gnu tools are much better and has the expected features) and the linux kernel is much more up to date on hardware support (Maybe not too relevant to most people but I remember FreeBSD announcing support for 64 CPUs recently, while Linux is already at 4096). Unless things have changed, the startup scripts are barely more advanced than DOS, and inflicting csh or tsch on people is evil given it is harmful. :) So for me, bsd is only a necessary evil to be used if linux won't run on the hardware, and the last time I had to resort to netbsd to get a machine running and doing useful stuff was about 18 years ago. I just can't find anything BSD does that linux doesn't or that BSD does better than linux. Unless you really don't like the GPL. -- Len Sorensen
On Thu, Sep 29, 2016 at 10:45:09AM -0400, Lennart Sorensen via talk wrote:
So for me, bsd is only a necessary evil to be used if linux won't run on the hardware, and the last time I had to resort to netbsd to get a machine running and doing useful stuff was about 18 years ago.
The *BSD software ecosystem lags behind Linux in hardware support; as far as I'm concerned, the ports/packages system is years behind Gentoo, Debian, Arch, and others; the communities are smaller and much less open. For all that... OpenBSD is where we get things like openssh from. Their packet-filtering system, PF, is a joy and a delight. The continuous code-auditing means that kernelspace *and* userspace programs that make up the system are often rewritten to get rid of cruft, and run cleanly and efficiently. They are a remarkable implementation of UNIX, without the ``better ideas'' such as systemd, and they are all phenomenally stable -- even more so than Linux. Granted, some of that comes from not having drivers for the latest hardware and innovations. But if clean and debugged code matter to you, if you want better deep ideas like privilege separation and default security, then I'd say there is a reason to have a look at OpenBSD. If you run a server, then you should definitely look at it. There's a reason why the internet ran on *BSD for so long, and why much of it still does. FreeBSD is the most common platform but has fewer distinctive features. NetBSD runs on almost anything, including toothbrushes, but is otherwise pretty plain vanilla. -- Peter King peter.king@utoronto.ca Department of Philosophy 170 St. George Street #521 The University of Toronto (416)-978-3311 ofc Toronto, ON M5R 2M8 CANADA http://individual.utoronto.ca/pking/ ========================================================================= GPG keyID 0x7587EC42 (2B14 A355 46BC 2A16 D0BC 36F5 1FE6 D32A 7587 EC42) gpg --keyserver pgp.mit.edu --recv-keys 7587EC42
On 09/29/2016 11:52 PM, Peter King via talk wrote:
On Thu, Sep 29, 2016 at 10:45:09AM -0400, Lennart Sorensen via talk wrote:
So for me, bsd is only a necessary evil to be used if linux won't run on the hardware, and the last time I had to resort to netbsd to get a machine running and doing useful stuff was about 18 years ago. This is the GTALinuxUserGroup...
In general I am not a fan of BSD and my experiences go back as far as the 80's Once I found linux and started downloading the floppy sets I never looked back. I have run into the argument of people claiming that some BSD variant is better than linux in some way or other and often when pressed the comparison is years old and based on a 0.99 version of the Linux kernel. Because of its dominant position Linux tends to vacuum up good ideas from all other operating systems.
The *BSD software ecosystem lags behind Linux in hardware support; as far as I'm concerned, the ports/packages system is years behind Gentoo, Debian, Arch, and others; the communities are smaller and much less open. In general linux packaging systems under linux have moved forward greatly over the years and for the most part it seems that the packaging system is what tends to be the first discriminator between distributions.
For all that...
OpenBSD is where we get things like openssh from. Their packet-filtering system, PF, is a joy and a delight. The continuous code-auditing means that kernelspace *and* userspace programs that make up the system are often rewritten to get rid of cruft, and run cleanly and efficiently. They are a
SSH and PF were developed in the days when linux was not a real option for people working in research and since then both features have been included in linux along with a lot of linux only development. I cannot comment much on the code auditing but that will be a small subset of the software that people commonly think of as the "system". For example I am sure the code auditing does not include things like X and the associated window managers which most people would think of as part of the "system". When someone installs Ubuntu they get a whole environment installed by default not just a kernel, package manager and some utilities(ala slackware et al on floppies).
remarkable implementation of UNIX, without the ``better ideas'' such as systemd, and they are all phenomenally stable -- even more so than Linux. Not sure why people have a hate on for systemd. It is a pain to learn a new way to manage your systems but it solves a number of problems and gets systems into a usable state faster in the face of startup problems. I curse systemd on a daily basis because my fingers know init but quite frankly having to wait 30 minutes for a system to boot up with init because some network connections need to time out is a major pain when its a critical system and the phones are all lit up. systemd removes the single threaded-ness of init and also provides a much better mechanism for dependency resolution.
Linux is also amazingly stable if you are willing to stay away from the bleeding edge. I am sure you will spend lots of time rebooting if you like to run the BSD nightly releases. I have a lot of Linux systems that I need to support that are well over 10 years old.
Granted, some of that comes from not having drivers for the latest hardware and innovations. But if clean and debugged code matter to you, if you want better deep ideas like privilege separation and default security, then I'd A number of that security ideas go back to the 80s. Sun was selling Compartment Mode workstations in the 90's Some very inserting work was done on secure operating systems at the UofT using the Turing programming language and a complete redevelopment and redesign of a unix compatible OS.
Linux now has selinux after a lot of work from the NSA and it incorporates ideas back to the 80s' and before. Linux has cgroups that are the basis of the container boom but it has a number of interesting abilities when used for complex network configurations.
say there is a reason to have a look at OpenBSD. If you run a server, then you should definitely look at it. There's a reason why the internet ran on *BSD for so long, and why much of it still does. I would kind of dispute that much of the internet runs on BSD. The guys at Cisco would be way more aggressive about refuting that. I still have the SUN information highway T-shirt but at the time I was working there I was running my ISP on linux. If you look at just about every home router appliance out there you will find it has a whole lot of GPL code built in with the kernel included. So by sheer numbers of computers BSD is going to be way down there. Then you have all the android devices.....
FreeBSD is the most common platform but has fewer distinctive features.
NetBSD runs on almost anything, including toothbrushes, but is otherwise pretty plain vanilla.
LUGians unite we can defeat the evil hoards of OS disbelievers ;) -- Alvin Starr || voice: (905)513-7688 Netvel Inc. || Cell: (416)806-0133 alvin@netvel.net ||
On Fri, Sep 30, 2016 at 9:28 AM, Alvin Starr via talk <talk@gtalug.org> wrote:
On 09/29/2016 11:52 PM, Peter King via talk wrote:
On Thu, Sep 29, 2016 at 10:45:09AM -0400, Lennart Sorensen via talk wrote: snip
Not sure why people have a hate on for systemd. It is a pain to learn a new way to manage your systems but it solves a number of problems and gets systems into a usable state faster in the face of startup problems. I curse systemd on a daily basis because my fingers know init but quite frankly having to wait 30 minutes for a system to boot up with init because some network connections need to time out is a major pain when its a critical system and the phones are all lit up. systemd removes the single threaded-ness of init and also provides a much better mechanism for dependency resolution. snip
Well - - - I can tell you why I find systemd a royal PITA. Systemd wants to be everything to everybody. That's astronomically difficult to do and what is in place today doesn't work half as well as it purports to. I have run into some of the issues which have resulted in a lot of hair pulling (hard when there's little left) in the process of resolving issues. I think that the original *nix thinking of doing one thing (at a time) and doing it well or better is my preferred solution. Part of the problem is that, even in linux, there are too many silos being built and not enough communication. I wonder if that is because most of the code writers are not really human communicators rather they are far better machine communicators? What say you? Dee
On Fri, Sep 30, 2016 at 10:47 AM, o1bigtenor via talk <talk@gtalug.org> wrote:
On Fri, Sep 30, 2016 at 9:28 AM, Alvin Starr via talk <talk@gtalug.org> wrote:
On 09/29/2016 11:52 PM, Peter King via talk wrote:
On Thu, Sep 29, 2016 at 10:45:09AM -0400, Lennart Sorensen via talk wrote: snip
Not sure why people have a hate on for systemd. It is a pain to learn a new way to manage your systems but it solves a number of problems and gets systems into a usable state faster in the face of startup problems. I curse systemd on a daily basis because my fingers know init but quite frankly having to wait 30 minutes for a system to boot up with init because some network connections need to time out is a major pain when its a critical system and the phones are all lit up. systemd removes the single threaded-ness of init and also provides a much better mechanism for dependency resolution. snip
Well - - - I can tell you why I find systemd a royal PITA. Systemd wants to be everything to everybody. That's astronomically difficult to do and what is in place today doesn't work half as well as it purports to. I have run into some of the issues which have resulted in a lot of hair pulling (hard when there's little left) in the process of resolving issues.
I am curious to know what some of these issues are. (Feel free to privately email me if you feel more comfortable with that). Systemd is not everything to everybody. It is a number of distinct binaries (doing that one thing at a time bit). As systemd keeps evolving, they hit limits of existing tools, and instead of waiting for them to catch up, they just rewrite and move on. Maybe we might actually end up with a plumbing layer unified across all distros.
I think that the original *nix thinking of doing one thing (at a time) and doing it well or better is my preferred solution. Part of the problem is that, even in linux, there are too many silos being built and not enough communication.
I wonder if that is because most of the code writers are not really human communicators rather they are far better machine communicators? What say you?
I feel that is unfair. Dhaval
On Fri, Sep 30, 2016 at 9:28 AM, Alvin Starr via talk <talk@gtalug.org> wrote:
On 09/29/2016 11:52 PM, Peter King via talk wrote:
On Thu, Sep 29, 2016 at 10:45:09AM -0400, Lennart Sorensen via talk wrote: snip Not sure why people have a hate on for systemd. It is a pain to learn a new way to manage your systems but it solves a number of problems and gets systems into a usable state faster in the face of startup problems. I curse systemd on a daily basis because my fingers know init but quite frankly having to wait 30 minutes for a system to boot up with init because some network connections need to time out is a major pain when its a critical system and the phones are all lit up. systemd removes the single threaded-ness of init and also provides a much better mechanism for dependency resolution. snip
Well - - - I can tell you why I find systemd a royal PITA. Systemd wants to be everything to everybody. That's astronomically difficult to do and what is in place today doesn't work half as well as it purports to. I have run into some of the issues which have resulted in a lot of hair pulling (hard when there's little left) in the process of resolving issues.
I think that the original *nix thinking of doing one thing (at a time) and doing it well or better is my preferred solution. Part of the problem is that, even in Well...
On 09/30/2016 10:47 AM, o1bigtenor wrote: the init process we are use to is not the original init process used by linux. Once again linux vacuumed up something, in this case the system-V init and tuned it with a whole batch of problems when it was first introduced. Now it works well but for the fact that there can be a huge number of init processes starting at system start up and the dependencies are not handled well. There was thought of using a make like system to process the dependencies but that would carry its own bag of problems.
linux, there are too many silos being built and not enough communication.
I wonder if that is because most of the code writers are not really human communicators rather they are far better machine communicators? What say you?
Dee . There is an old programmer axiom. "If it was hard to write it should be hard to understand"
But yes. As a rule FOSS documentation sucks because it takes lots of time and often more time than the original work. Combine that with people using the support and documentation as a way to get paid for the project. You end up with bad documents by nature and by design. -- Alvin Starr || voice: (905)513-7688 Netvel Inc. || Cell: (416)806-0133 alvin@netvel.net ||
<snip systemd automation heuristics>
I wonder if that is because most of the code writers are not really human communicators rather they are far better machine communicators? What say you?
We have to understand Linux userland is enormous now. The learning curve for network operator's, is larger than those who develop systems, ability to communicate the manner in which systems are changing, across the board, so to speak. Imagine this planets surface is the planar board and all the Ethernet connections are tracings and you can see the scope of the scale of current development. Systemd and systemctl are efforts aimed at large scale networking and enterprise. Just as SElinux is aimed at large scale security in enterprise. This environment is often referred to as an ecosystem, it's truly only a mechosystem, engineered by the weakest link, the meatware.
Dee --- Talk Mailing List talk@gtalug.org https://gtalug.org/mailman/listinfo/talk
Russell Sent from mobile.
On Thu, Sep 29, 2016 at 11:52:59PM -0400, Peter King via talk wrote:
On Thu, Sep 29, 2016 at 10:45:09AM -0400, Lennart Sorensen via talk wrote:
So for me, bsd is only a necessary evil to be used if linux won't run on the hardware, and the last time I had to resort to netbsd to get a machine running and doing useful stuff was about 18 years ago.
The *BSD software ecosystem lags behind Linux in hardware support; as far as I'm concerned, the ports/packages system is years behind Gentoo, Debian, Arch, and others; the communities are smaller and much less open.
Oh you had to bring up ports. I had nicely forgotten about the existence of that. :(
For all that...
OpenBSD is where we get things like openssh from. Their packet-filtering system, PF, is a joy and a delight. The continuous code-auditing means that kernelspace *and* userspace programs that make up the system are often rewritten to get rid of cruft, and run cleanly and efficiently. They are a remarkable implementation of UNIX, without the ``better ideas'' such as systemd, and they are all phenomenally stable -- even more so than Linux. Granted, some of that comes from not having drivers for the latest hardware and innovations. But if clean and debugged code matter to you, if you want better deep ideas like privilege separation and default security, then I'd say there is a reason to have a look at OpenBSD. If you run a server, then you should definitely look at it. There's a reason why the internet ran on *BSD for so long, and why much of it still does.
I think a lot more runs on Linux these days, and it seems you don't need that high end a server before you exceed the capabilities of BSD to even support these days. I see the openbsd release notes for version 5.9 says that installing to a GPT partitioned disk is new in this release, from this year. So they didn't support installing on UEFI systems until this year? No support for installing on larger than 2TB disks before now? Really?
FreeBSD is the most common platform but has fewer distinctive features.
NetBSD runs on almost anything, including toothbrushes, but is otherwise pretty plain vanilla.
-- Len Sorensen
William Park via talk wrote:
To those who knows/uses both BSD and Linux... Should I learn BSD, and which one?
If you read to HackNews we are currently in the systemd apocalypse and Linux's user base is shrinking every day and good ethical people are moving to BSD to the warm embrace of init. Without sarcasm, learning another system is always a good idea because it gives you more insight on how others work. As an example I would have never been able to understand how Google's open source Python code worked without some knowledge of Java. -- Have an awesome day, and happy open-sourcing :-).
On 29 September 2016 at 11:00, Myles Braithwaite via talk <talk@gtalug.org> wrote:
William Park via talk wrote:
To those who knows/uses both BSD and Linux... Should I learn BSD, and which one?
If you read to HackNews we are currently in the systemd apocalypse and Linux's user base is shrinking every day and good ethical people are moving to BSD to the warm embrace of init.
There is something to be worried about there... Though the answer seems unlikely to keep heads in the sand, as the reasons that systemd emerged include some pretty valid ones....
Without sarcasm, learning another system is always a good idea because it gives you more insight on how others work. As an example I would have never been able to understand how Google's open source Python code worked without some knowledge of Java.
There's a Debian port to FreeBSD <https://www.debian.org/ports/kfreebsd-gnu/>, so you could have a mostly-GNU userspace that presumably lacks systemd. I'm occasionally attracted to take a peek at Dragonfly BSD, as it has been trying to do some substantial reimplementations of some of the internals with particular view to improving performance and supporting clustering. The HAMMER filesystem is one of the interesting bits; some data deduplication capabilities, and a BSD flavour on the "advanced" stuff like snapshotting, journalling, et al. Mind you, the idea hasn't been interesting enough to lead to my having any systems running such. I considered throwing Debian/kfreebsd onto my media box, but the curious inability to get it to boot off CDROM wound up curbing experimentation. (I wound up using PXE to pull a recent Debian image off another machine; "thanks Scott for your PXE talks!!!") -- When confronted by a difficult problem, solve it by reducing it to the question, "How would the Lone Ranger handle this?"
I have been using Linux (Debian primarily),FreeBSD, OpenBSD and Solaris for about 15 years now. IMHO with Linux, generally things just work, drivers, dependencies, X etc. The kernel comes with more features and capabilities and is generally more useful and practical. In contrast, FreeBSD and OpenBSD take longer to introduce features to the base distribution and kernel. Case in point, FreeBSD and NetBSD both have raspberry PI support. The GPIO capability on FreeBSD and NetBSD is pretty poor as compared to Rasbian. OpenBSD hasn't even started to port to Raspberry PI yet. The one thing that OpenBSD has that is much better than all the other platforms is the PF firewall. The variant of PF included with FreeBSD is a fork that has not kept up with the advances on OpenBSD. I had hoped with Apple's use of the FreeBSD kernel in OSX and its subsequent acquisition of a lot of very talented BSD folks that there would have been a huge contribution back to the open source community. FWIW, I don't care too much about supporting 64 CPUs. Both FreeBSD and Linux had at one point BGL issues that affected SMP performance. However over the last 10 years most Linux distributions have become like the "kitchen sink" in that they throw everything in (both in the Kernel and userland). Installation sizes are incredibly bloated now. I work with a lot of enterprise sized companies (>1000 servers). I don't see any other FOSS OSes other than Linux. If you interest is widening your scope for work, then IMHO learn Solaris. If you want to have a lot of fun and turn a few hairs grey, try PF on OpenBSD. On Thu, Sep 29, 2016 at 4:14 PM, Christopher Browne via talk < talk@gtalug.org> wrote:
On 29 September 2016 at 11:00, Myles Braithwaite via talk <talk@gtalug.org> wrote:
William Park via talk wrote:
To those who knows/uses both BSD and Linux... Should I learn BSD, and which one?
If you read to HackNews we are currently in the systemd apocalypse and Linux's user base is shrinking every day and good ethical people are moving to BSD to the warm embrace of init.
There is something to be worried about there...
Though the answer seems unlikely to keep heads in the sand, as the reasons that systemd emerged include some pretty valid ones....
Without sarcasm, learning another system is always a good idea because it gives you more insight on how others work. As an example I would have never been able to understand how Google's open source Python code worked without some knowledge of Java.
There's a Debian port to FreeBSD <https://www.debian.org/ports/kfreebsd-gnu/>, so you could have a mostly-GNU userspace that presumably lacks systemd.
I'm occasionally attracted to take a peek at Dragonfly BSD, as it has been trying to do some substantial reimplementations of some of the internals with particular view to improving performance and supporting clustering. The HAMMER filesystem is one of the interesting bits; some data deduplication capabilities, and a BSD flavour on the "advanced" stuff like snapshotting, journalling, et al.
Mind you, the idea hasn't been interesting enough to lead to my having any systems running such. I considered throwing Debian/kfreebsd onto my media box, but the curious inability to get it to boot off CDROM wound up curbing experimentation. (I wound up using PXE to pull a recent Debian image off another machine; "thanks Scott for your PXE talks!!!") -- When confronted by a difficult problem, solve it by reducing it to the question, "How would the Lone Ranger handle this?" --- Talk Mailing List talk@gtalug.org https://gtalug.org/mailman/listinfo/talk
Greetings To GTALUG Talk, ----- Original Message ----- From: "Christopher Browne via talk" <talk@gtalug.org> To: <myles@gtalug.org>; "Myles Braithwaite" <me@mylesbraithwaite.com>; "GTALUG Talk" <talk@gtalug.org> Sent: Thursday, September 29, 2016 4:14 PM Subject: Re: [GTALUG] curious... Linux vs BSD ? <snip>
I'm occasionally attracted to take a peek at Dragonfly BSD, as it has been trying to do some substantial reimplementations of some of the internals with particular view to improving performance and supporting clustering. The HAMMER filesystem is one of the interesting bits; some data deduplication capabilities, and a BSD flavour on the "advanced" stuff like snapshotting, journalling, et al.
I'm also intrigued by DragonFlyBSD (dfly). Mainly for the same two reasons that Chistopher Browne cites: 1. great speed, 2. robust HAMMER file system (HAMMER1). Although my project to use dfly has been on hold due to other priorities, I did install dfly on a QEMU / KVM virtual server at www.elastichosts.com. Haven't done much with dfly there yet. The plan is to use dfly for a website (php, postgresql, nginx). I am subscribed to the dfly general discussion list. Mostly just lurking. Very encouraged to see the dfly team respond well, to calls for help with dfly failures, from another list participant who is running an app under dfly on a VM. More than one fix to the dfly virtio drivers has been delivered. It seems that most dfly servers run on bare metal. But I have argued on the dfly list for good dfly support also for QEMU / KVM VMs as they are a good entry point for budget-constrained projects based on dfly. There also are a few people trying to use dfly as a desktop o/s. However, the main thrust of dfly seems to be in the server arena. Steve
participants (12)
-
Alvin Starr -
Ansar Mohammed -
Christopher Browne -
Dhaval Giani -
lsorense＠csclub.uwaterloo.ca -
Myles Braithwaite -
o1bigtenor -
Peter King -
Russell Reiter -
Steve Petrie, P.Eng. -
Stewart C. Russell -
William Park