Gnu/Linux Considered Harmful

[edy]

While various Linux distros are far from perfect, including Ubuntu (which is the main distro I use, http://ubuntustudio.org/, and love it), I can tell you that the last few years of vastly-improved installation, configuration and GUI tools to manage the systems has *GREATLY* improved the reach of Linux to your average "n00b" (like myself) who would have otherwise NEVER even thought of using anything other than Windows!

Also, the availability of so much free open source software now to handle almost everything I was doing on Windows made the jump quite easy, and for those rare instances that I still need to run some legacy Windows software I can either use WINE or have VMWare inside my Ubuntu machine when I need it.

Another bonus is that I can often run Linux on a wide range of hardware of different power, from RasPi through to the most powerful servers, and carry my knowledge across them... and also put Linux on machines that are older and would otherwise croak to a grinding halt trying to run any functional modern version of Windows. I can find a current (but more simplified) Linux distro that takes a fraction of the RAM and HD space and still run fast and secure and be updated, and put it on 5-10 year old hardware no problem and be fully functional!   

Linux has opened up the eyes of users who have had really no other practical alternative to Windows. Nothing is perfect, but now users have a CHOICE. If you think about it, what operating system alternatives do PC users have? For that matter, what options to Mac users have? After Windows and MacOS, the next choice happens to Linux... and a number of distros to suit various tastes, most if not all free, and now far easier and within reach of anyone who wants to install and configure it as the support for hardware has improved tremendously (some big vendors are finally starting to contribute). The use of liveUSB's also lets you try the systems without installing, and you can even configure them with Persistent Storage.

Yes, the current crop of Linux OS's may not appeal to the more puritan Unix users. They are often a cobbled patchwork of software and forks (as would be expected in an "open" environment) and there can be inconsistent experiences between Linux systems. But back to the title of the original thread... is it considered "harmful"? In what way? All I see is that computer and software enthusiasts have work hard over decades to create a robust and as much as they could well-engineered OS (often on a volunteer basis with no pay involved) to create an ALTERNATIVE to the monopoly of Microsoft on the PC world. This can only be GOOD.

And by the way, I can share with you plenty of issues with Windows 10 working in a business environment that have caused countless headaches with absolutely no obvious way to track down the bugs. It was only after searching through countless pages and finding obscure support blogs that you find idiosyncrasies creeping in by each Windows update that messes up existing working systems! Linux is not perfect either, but it offers me control and customization that I could never imagine ever doing in Windows, and finally a CHOICE which ultimately gives me FREEDOM.

Again, I can't vouch for anyone doing more sophisticated stuff like compiling thousands of lines of ported code, development, or those using specialized software for which no Linux options exist... But for a VAST MAJORITY of users who live in the Windows world and do consider themselves "Power Users", Linux is perfectly suitable and has the tools to enable them to be very efficient and productive.

[bd139]

I was thinking about this earlier and I think it goes like this now (apologies to those of us without dicks):

1. User gets windows and is punched in the dick.
2. User gets fed up of being punched in dick and someone tells them about Linux.
3. User installs Linux and is punched in the dick.
4. User installs windows again because at least they knew roughly what they were doing while being punched in the dick.
5. User gets fed up of being punched in dick and someone tells them about Apple
6. User goes to Apple store and gets mugged and punched in the dick.

Common problem is user is punched in the dick wherever they turn.

Winning vendor needs to write something which doesn’t punch you in the dick.

[SiliconWizard]

Common problem is user is punched in the dick wherever they turn.

 

But we still manage to do useful stuff out of all those platforms, at least most of us. And all those willing to do better and thinking they have all the right ideas usually end up either with something like Linux (at best), or even like BeOS (does anyone here remember BeOS? )

One big advantage of Linux is that you can always make your own distribution (even if that's time-consuming) that fits your own requirements (which are often different from your neighbour's), whereas it's pretty much impossible with Windows or MacOS.

[bd139]

There’s an illusion of productivity based on “today isn’t as bad as that really bad day we had”.

I remember BeOS. That was where we should have gone.

Apart from C++

[SiliconWizard]

There's an old saying stating that a good developer makes approximately 1 bug every 800 LOC.

After typing in the source code?
After compiling?
After testing?
After delivery to customemers?

Obviously that's just an old saying, something I used to hear more than 20 years ago in some circles. Not something that has any scientific value.

It was kind of a rule-of-thumb thing that appeared not that far away from real-world facts at the time, as I remember. Of course you first have to define what you consider a bug, and what you consider a good developer. Here since this was centered on individual developers, you can't consider it in convoluted settings with many stages of development and many different people involved in testing, "continuous delivery" and all this.
Bugs considered here would be bugs that eluded the developer themselves, thus not the bugs they inevitably make during developing and that they catch while they develop, including compiling and their own testing.

Whereas the relevance of this kind of metric is highly debatable, it has been studied by many people and some have even concluded that the programming language, as well as the development method, influenced this metric only marginally, something that with experience, I tend to agree with (but obviously something that all the proponents of software silver bullets will fiercely deny).

[rhb]

What made Unix so successful, was a knowledgeable user could easily do anything they wanted.  The amount of detail that you had to no from memory was small and discoverability was simple.  The only thing that I know of  that improved on that is Plan 9 which introduced the concept of configuring devices via ASCII files.  I got a 3.5" floppy of the first release at Usenix '95 in NOLA.  An operating system, windowing system, editor, full UNICODE support and the basic command line tools in less than 1.44 MB.

I now have the needed mtd kernel modules placed in the appropriate places, but when I attempt

"modprobe mtdram total_size=64000"

I get a message saying it can't find the module.  So I am reduced to using strace(1) to find out where this idiot system is looking.

Can I make this work?  If it does work on CentOS, yes.  But the labor is ludicrous.  And I read some man pages which said something to the effect:  "This is a trivial program for doing XYZ."  Absolutely *no* useful information supplied.

The *only* reason I ever use Gnu/Linux is because the software I need to use cannot be run anywhere else.

[malagas_on_fire]

When it comes to compilation process it is better to have some documentation , specially when it comes to kernel. The doc folder should be present .

The menuconfig has keys for search fields (/) and a help (?) on the menus. It has grown a lot and it requires some reading from the Documentation folder. It is best to work the kernel with a repository to  track changes.

The main key is the documentation and information provided on the kernel issues, software issues that might be found in daily use to better improve them.

About being punched well you could opt not having a banal computer , maybe a chromebook  Just kidding

[Monkeh]

So I am reduced to using strace(1) to find out where this idiot system is looking.

Or you could simply look up where modules are stored.

Have a freebee: /lib/modules/$(uname -r)

[bd139]

Yes. Kernel modules are really easy to compile and install (and write!). Literally a 4 line cut and paste makefile.

I repeat what I've said earlier probably incoherently that the Linux kernel is a nice bit of tech. Unfortunately it's a figurehead on a ship of fools on an ocean of diarrhea. It's the userland that sucks.

[malagas_on_fire]

That is true when it comes for the user interface and i think i've comment about it. the simple interfaces like LXDE and XFCE works good since they have been ported to the SBC world but they offer less functions compared to the windows gui. Sometimes the linux gui crashes maybe with samba folders and / or davfs, .. it is more the application Thunar and then you have to save you're work, go to CLI, restart graphic services. if you can't logout with GUI....

It is way better than some years ago the GUI but not as close on the windows . Again  linux is more like to be a learning OS .

[bd139]

I never got those weird ass virtual filesystems in file managers in desktop environments. GVFS was a crime. The damn OS has a perfectly good VFS layer. Why build another one?!?!?! Crack heads.

(Windows connection to phones etc is just as bad)

[rhb]

So I am reduced to using strace(1) to find out where this idiot system is looking.

Or you could simply look up where modules are stored.

Have a freebee: /lib/modules/$(uname -r)

*That* I already knew.  They were already installed  where they belong.  The question is what other files is modprobe reading.  Eventually I learned from dmesg that the modules were considered tainted.

I finally concluded that the CentOS 7 mtd modules are grossly out of date.  I'm currently installing Debian 9.3 as I am fairly confident they work on that.   I have a system with a trayless SATA drive mount and a tray style PATA drive mount.  It's a great use for old drives.  So a 1 TB drive  leftover from upgrading my my Solaris 10 system will become my work system for memory technology devices.  And Centos 7.4 will be my Vivado system.  I tried VirtualBox VMs, but the Illumos/OpenIndiana USB support is not very good.

Until a few years ago, I refused to use Windows outside of access to the mail system at work.  And I did not really use Linux for anything other than testing new releases of some large seismic processing packages I supported for many years.  But there are a lot of electronics engineering applications which only run on Windows.  So having picked up 3 Z400s for $100 each I configured one so I can boot off of any of the now rather large collection of old hard drives I have accumulated.

My primary systems are a Solaris 10 u8 running ZFS on a 3 disk RAIDZ1 array and an OpenIndiana Hipster system set up the same way.  For security reasons the S10 system is *never* connected to the internet even though it is sitting on a NAT.  I have way too much work invested in configuring that for doing development and maintenance work on large software packages.

Rather sadly, my recent experience with CentOS aka Red Hat makes Win 7 Pro look good.  I *never* thought I'd ever make such a comment.  Of course, everything since Win 7 is a nightmare.

As a general response to some comments.  If you personally do not *start* solving a problem with an unfamiliar program by tracing the system calls or know people who do, you can't begin to appreciate the skill level that implies.  The primary reason I'm having problems with this stuff are two fold:

Gnu/Linux has turned into a terminally complex and inconsistent pile of crap

I've not been doing serious software work for 7-8 years, so I am *very* rusty

The fact that Gnu/Linux documentation is a pile of conflicting, poorly written and inaccurate crap is only an issue for me because I've not done any serious software and system administration work in a long time.  And if you don't do it, it all just fades away. 

The best sys admin I have ever met is now a senior manager with AT&T consulting managing projects all around the world.  Whenever we talk, the subject of our having been reduced to what we consider incompetent by lack of practice always comes up.  I'm a bit better than he is, as he has not touched a Unix system in 10-15 years.  But neither of us is anything close to what we were when we worked together and were in top form.

[Monkeh]

As a general response to some comments.  If you personally do not *start* solving a problem with an unfamiliar program by tracing the system calls or know people who do, you can't begin to appreciate the skill level that implies.

And how did that work out for you?

Eventually I learned from dmesg that the modules were considered tainted.

Right, you chased your tail until you looked at a standard resource for kernel messages.

And yes, CentOS is grossly out of date - if that's not apparent from the fact that it was released nearly four and a half years ago and uses a five and a half year old kernel, I begin to understand your problem.

[rhb]

As a general response to some comments.  If you personally do not *start* solving a problem with an unfamiliar program by tracing the system calls or know people who do, you can't begin to appreciate the skill level that implies.

And how did that work out for you?

I'd say rather well.  I was paid oil industry Stanford PhD day rates despite not having gotten my doctorate.  I have several  very close friends in that category.  So I have a very good idea of what the pay scales are.  A fairly important detail if you work as a contractor on 3-6 year contracts.  I did not look for work.  It looked for me by name.

Eventually I learned from dmesg that the modules were considered tainted.

Right, you chased your tail until you looked at a standard resource for kernel messages.

And yes, CentOS is grossly out of date - if that's not apparent from the fact that it was released nearly four and a half years ago and uses a five and a half year old kernel, I begin to understand your problem.

I did not choose CentOS.  Xilinx did.  As for the "standard resource" I am quite aware of it.  But it seems a substantial number of the Gnu/Linux crowd are not.  So it is never clear *if* there is an error message logged, or *where* it is logged.
 
As a bit of perspective, this is a partial list of *nix systems I have worked on.  I'm sure there were some others, but these are what I can recall at the moment.

Unix OS variants or clones:

Minix
Coherent
SGI Irix
IBM AIX
Intergraph CLIX ( very primitive  Sys V on the Clipper chip, the CPU not the encryption chip)
HP "snakes" series
DEC Ultrix
Sun 386i ( rebranded Interactive System V)
SunOS 4.x
Solaris
Intel i386 Hypercube
Intel i860 Hypercube
Evans and Sutherland ?
Alliant ?
FreeBDS
OpenBSD
DEC/Compaq Tru64 (aka OSF)
Mac OS X
Slackware
Mandrake
Red Hat
Suse
Fedora
CentOS
Debian
Ubuntu

Having made a few million lines of old code work on any arbitrary set of these  I don't have anything else to prove.

The above does not include non-Unix systems like VMS, MVS, VM/CMS, Perkin-Elmer, MUSIC and several others.  I ran a VAX for 3 years and tuned it to run at 100% CPU utilization for months at a time running batch jobs *and* providing the interactive  response of an idle system.  I've never come across anyone else who figured out how to do that.

It is *very* expensive to qualify an OS update in a large corporate environment with a large number of 3rd party packages..  The applications oftn cost $100K per seat.  So RHEW moves *very* slowly with great caution.  I've had a front row seat to IT meltdowns because of version updates that cost hundreds of thousands of dollars per day for several days.  Management gets a bit excited when such things happen.

I was once asked to hep resolve some problems with a reservoir simulation package for which my client paid $80K/yr for support.  After a week or two of idiocy, the vendor support guy said, "If you get it working, please send me the changes to I can give them to the other customers."  I did not work for the reservoir guys and was just doing them a favor.  So at that point, I reported it to my supervisor and stopped work on it.  In the internal billing structure I was working for free.  I was getting paid, but out of someone else's budget.

BTW It's pretty clear you do *not* recognize the allusion in the title.

At his point, I think all of the interesting and enlightening posts are probably done.  My thanks to those who understood the issues for their comments.  Some people brought up points I either did not know or had not considered.  But it's getting unpleasant, so time to call a halt.

[Monkeh]

As a general response to some comments.  If you personally do not *start* solving a problem with an unfamiliar program by tracing the system calls or know people who do, you can't begin to appreciate the skill level that implies.

And how did that work out for you?

I'd say rather well.  I was paid oil industry Stanford PhD day rates despite not having gotten my doctorate.  I have several  very close friends in that category.  So I have a very good idea of what the pay scales are.  A fairly important detail if you work as a contractor on 3-6 year contracts.  I did not look for work.  It looked for me by name.

That's not what I asked and you know it. But keep on blowing your horn.

At his point, I think all of the interesting and enlightening posts are probably done.  My thanks to those who understood the issues for their comments.  Some people brought up points I either did not know or had not considered.  But it's getting unpleasant, so time to call a halt.

You do have some valid points, but you mix it in with bias, ageism, and arrogance.

So yes, I'm out, carry on.

[ebastler]

At his point, I think all of the interesting and enlightening posts are probably done.  My thanks to those who understood the issues for their comments.  Some people brought up points I either did not know or had not considered.  But it's getting unpleasant, so time to call a halt.

The main thing this thread has done for me is to confirm an apparently universal human tendency in tech:

• Those innovations which came about (or which we learned about) during our formative years, we consider the best thing since sliced bread. If you are in the right age bracket, these might include the original Unix systems and structured programming (no GOTOs please).
  
• Those innovations which come about later, when we are more settled, we tend to consider newfangled crap. Depending on your age, these might include object-oriented programming or Gnu/Linux, for example.

It's hard to impossible to rid oneself of such notions. I am certainly guilty of them myself. But it helps to be self-aware, and not assume that one owns the absolute truth...

[Marco]

The only person I know of who is actively addressing this is Andy Tanebaum and the Minix 3 project.

That's not really true, Linux's process and driver isolation slowly lumbers forward. They just have a lot more baggage to move in the process. For most of Linux's existence there weren't even no execute bits in the page table and there was a 1 bit address space ID for the TLB on their main platform. Linux was designed for the architecture it was given.

[borjam]

• Those innovations which came about (or which we learned about) during our formative years, we consider the best thing since sliced bread. If you are in the right age bracket, these might include the original Unix systems and structured programming (no GOTOs please).
  

I strongly disagree.

The fact that we still stick to an operating system born in the early 70's reflects the disgraceful state of operating systems research and its adoption by the industry, which as led to systems research becoming irrelevant.

This article by David Pressotto was published in 2000.
https://www.cse.iitb.ac.in/infolab/Data/utah2000.pdf

I don´t claim any of these ideas as mine of course, but I really agree with him.

• Those innovations which come about later, when we are more settled, we tend to consider newfangled crap. Depending on your age, these might include object-oriented programming or Gnu/Linux, for example.

Problem is, GNU Linux is NO innovation. In several areas it went backwards because of the ignorance or disdain for history or "grumpu old greybeards" by many of its contributors. It has been hit really hard by the obsession with software licenses. I can agree with some of the goals of the Free Software Foundation but you can hardly satisfy two different goals with an operating system. Either you get a briliant, useful, performant, cutting edge operating system or a way to support your political cause. Both? I think it's not possible.

I think the GNU project has brought a lot of good helping fuel the open source movement. They haven't achieved their final goals probably, but I don't think they were so good anywyay.

OOP. It's a briliant intellectual creation and it can be incredibly helpful in tackling software complexity. But there is more to software complexity than mere information hiding, and some of the problem is not actually tackled, but swept under the rug. So it's not the right solution for every problem and a good OO design is much harder to do than it's apparent.

Comparing OOP (a software paradigm) with a particular program (Linux) is like comparing apples to siphonophores.

[malagas_on_fire]

(Windows connection to phones etc is just as bad)

Except if it was a windows phone / mobile which work great in both Linux and Windows. Also Windows phone itself was very good, poorly anounced although Nowadays it's all using the MTP driver and the ADB driver support has improved a lot.

[Edit]
About being harmfull it depends the purposes on what Gnu/Linux is used or interpreted.  The most concerning thing may be the fragmentation on distribuitions ... there are so many... derivates..   

[bd139]

Windows phone was great. If they (a) ironed out all the bugs and (b) as per all MS products, didn't suffer from severe schizophrenia

[vtwin@cox.net]

And yes, CentOS is grossly out of date - if that's not apparent from the fact that it was released nearly four and a half years ago and uses a five and a half year old kernel, I begin to understand your problem.

This recently bit me in the ass (this week.) I use CentOS principally as I have licensed software at work which, should I require technical support on, the company will not even speak to me unless I am using one of their "supported" distributions, of which RHEL is one (learned the hard way a while back when I needed support and was running Fedora Core at the time). Folks at work are too cheap to pay for a software update contract ("Linux? Isn't that free? Why do I have to pay Redhat for updates?"). Thus, I tend to use CentOS for my personal stuff too, so I'm not trying to remember two different sets of configuration tools, etc.

Recently at home I had a dell server, which acted as my principal backup server (Windows) and internet router/firewall (CentOS HyperV), STB. Rather than replacing it with another large server, I opted to "go small", so I picked up a couple of Zotac ZBoxes, one for my router and the other to connect my 12TB backup drive to (the latter requiring a little bit of a hack to connect, mainly an external case with power supply for the drive with a SATA data cable running from it into the Zotac via a small "portal" I drilled into the case, to connect directly to the internal SATA connector, so the unit thinks it is a 12TB internal drive)

Well, booting CentOS on the ZBox was painfully slow. And I mean painfully slow. As in an hour to boot a minimum install thumbdrive just to get to the installation menu. On the other hand, I could boot a Debian stretch install image and have a built-out system in 15 minutes. I really didn't want to convert over to Stretch, as I have(had) my CentOS router configured the way I wanted it, and have all the config files ready to install on a new build.... compared to trying to "learn" where Debian puts everything to get it the way I wanted. My wife (IT auditor) works from home quite a bit and requires internet access for work, so downtime is something to be avoided if I do not want to sleep on the couch

After much research, turns out the problem is w/ the kernel (3) used by CentOS, and using a mainline kernel (4) off the elrepo repository fixed the problem. Now the ZBox flies (with the 4 kernel). Easily keeps up with my 300MB download speed from my ISP.

So, yeah, CentOS is a bit dated (like me!) I guess. I suppose I should start investing the time to familiarize myself with some "newer" distributions.

[bd139]

Yes it gets worse with centos as well because you have to pull EPEL in to do anything useful and half the shit in there is just broken.

[ebastler]

This article by David Pressotto was published in 2000.
https://www.cse.iitb.ac.in/infolab/Data/utah2000.pdf
I don´t claim any of these ideas as mine of course, but I really agree with him.

I would argue that this article is in fact another example for my (yes, overly generalized and simplified) statements:

Presotto Rob Pike believes that the research done in "his" days was relevant, and that the more recent directions of the computer science and IT world are newfangled crap. Designing new operating systems is what the world needs in his view, that whole usability stuff is irrelevant.


Edit: Changed the attribution. I hope we are talking about the same article? The one you linked to is not by D.P.

[sokoloff]

As a general response to some comments.  If you personally do not *start* solving a problem with an unfamiliar program by tracing the system calls or know people who do, you can't begin to appreciate the skill level that implies.  The primary reason I'm having problems with this stuff are two fold:

I don't think *starting* with strace/dtrace is a remotely reasonable course of action. Start with man pages, start by reading log messages/eventvwr logs, start with a Google search. Start almost anywhere but tracing syscalls.

If I took my sick kid into a pediatrician or even ER and the beginning of diagnosis was a CT-scan, I sure wouldn't be thinking, "Wow, such skillful doctors!" Maybe starting with a thermometer and stethoscope is a lot smart course of action...

[ebastler]

Comparing OOP (a software paradigm) with a particular program (Linux) is like comparing apples to siphonophores.

Huh? I used them as two completely independent examples of older vs. more recent innovations in computer science. Unix/Linux on one hand, and structured/object oriented programming on the other hand. Of course there is no point comparing Linux vs. OOP, and I didn't compare them.