Author Topic: Gnu/Linux Considered Harmful  (Read 45170 times)

0 Members and 1 Guest are viewing this topic.

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23018
  • Country: gb
Re: Gnu/Linux Considered Harmful
« Reply #175 on: November 16, 2018, 02:17:53 pm »
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.

That's not what he suggested. He suggested systems research was done, which it pretty much is. Turns out the application of the technology, not the technology was the big thing and he was right.

As for "recent directions" and "newfangled crap" Rob Pike (and Ken Thompson) are currently looking after Go at Google...
 

Online ebastler

  • Super Contributor
  • ***
  • Posts: 6202
  • Country: de
Re: Gnu/Linux Considered Harmful
« Reply #176 on: November 16, 2018, 02:46:30 pm »
That's not what he suggested. He suggested systems research was done, which it pretty much is. Turns out the application of the technology, not the technology was the big thing and he was right.

As for "recent directions" and "newfangled crap" Rob Pike (and Ken Thompson) are currently looking after Go at Google...

To your first point: No, I don't think Rob's drift was that there is nothing more to be done in system software research. When he says that "systems software research has become irrelevant", he does not mean that there is nothing left to do, but that software researchers have been focusing on other things, which he regrets.

Regarding the "newfangled" part, you have a point. Maybe Rob has managed to avoid the second one of my favorite fallacies, and is still open to new stuff.  ;)
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #177 on: November 16, 2018, 04:48:34 pm »
I strongly agree with Rob Pike.  Thanks for the link.

Obviously one reads whatever documentation is available.  I was referring to what you do when something doesn't work.  Especially if it fails silently.  The man pages on CentOS are pathetic.  Lots of commands have no man page at all.

 

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23018
  • Country: gb
Re: Gnu/Linux Considered Harmful
« Reply #178 on: November 16, 2018, 05:20:42 pm »
I think they assume you can just google stuff.

Which is no good when your network connection is down, you're in the middle of nowhere with no phone reception. Been there. Probably already said that in this thead  :-DD

OpenBSD is the only thing that doesn't scare me in that department. You can run just the base system and never be SOL.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #179 on: November 16, 2018, 08:46:47 pm »
I installed Debian 9.3 on a hard drive and life is much better now.  I'm still stuck with CentOS for Vivado and possibly the Altera tools.  I don't recall what those support.  I was hoping to avoid having more filesystems to root through looking for stuff I did last year.

It will be interesting to see if the quality of RH improves now that IBM owns it.  I was helping a friend set up a compute cluster circa 2000 and RH support was pathetic.  I was on hold forever and then the person didn't know as much about it as I did.

The  reason I went straight to strace(1) was to see if modprobe was looking in a configuration file.  I've seen far too many instances of you had to fiddle someplace that wasn't documented and when it failed to find what it was looking for exited without issuing a meaningful message.  In this case though the real issue is the antiquity of CentOS 7.  But as for how that worked out for me, it did *exactly* what I wanted to do.
 

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23018
  • Country: gb
Re: Gnu/Linux Considered Harmful
« Reply #180 on: November 16, 2018, 09:49:24 pm »
It will be interesting to see if the quality of RH improves now that IBM owns it.

The only thing that will improve is the number of consultants and there will be Notes bundled.
 

Offline borjam

  • Supporter
  • ****
  • Posts: 908
  • Country: es
  • EA2EKH
Re: Gnu/Linux Considered Harmful
« Reply #181 on: November 17, 2018, 10:06:42 am »
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.
I read it when it was published, but the main issue was, systems software research had been rendered irrelevant not because it was a “complete” field, but because the market was happy with old and even crappy solutions and nobody wanted to bother adopting anything new.

Quote
Edit: Changed the attribution. I hope we are talking about the same article? The one you linked to is not by D.P.
Yes, my fault. Both were members of the same research group at Bell Labs and both coauthored several papers about the Plan 9 and Inferno operating systems.

 

Offline borjam

  • Supporter
  • ****
  • Posts: 908
  • Country: es
  • EA2EKH
Re: Gnu/Linux Considered Harmful
« Reply #182 on: November 17, 2018, 10:10:36 am »
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.
Sorry, my answer was poorly worded. Anyway, Linux is not an innovation. At least in my close environment, where people favors Linux Unix strongly, Linux was received really eagerly. At the time you had SCO Unix (which was a mess) to run on cheap PC class computers. And the disappointment was huge, that is in my opinion the reason why Linux hasn’t been really accepted by many, including myself.

« Last Edit: November 17, 2018, 06:48:48 pm by borjam »
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #183 on: November 17, 2018, 12:58:26 pm »
[
I read it when it was published, but the main issue was, systems software research had been rendered irrelevant not because it was a “complete” field, but because the market was happy with old and even crappy solutions and nobody wanted to bother adopting anything new.

Quote
Edit: Changed the attribution. I hope we are talking about the same article? The one you linked to is not by D.P.
Yes, my fault. Both were members of the same research group at Bell Labs and both coauthored several papers about the Plan 9 and Inferno operating systems.

This is the first time I've come across anyone mentioning either of those. I was particularly attracted to Plan 9 for large (10,000+ users) environments. I chatted with Ritchie about that aspect at Usenix '95.  They never did the work to really provide good support for large number of users.  The focus shifted to the set top box and Inferno.

Of course, I agree with Andy Tanenbaum's comment that Linus would have flunked the Minix course.  So far as I know Andy is the sole person still actively doing OS research.  As Pike notes in the linked paper, making Plan 9 compatible with various standards was most of the work.

The real problem is that the amount of code that needs to be usable on *any* replacement for Windows, *nix or MVS is so huge that the inertia is overwhelming.

I'll probably use Solaris 10 until I die for one simple reason.  The Sun/Forte dbx is the only debugger that will let you do a "print cos(a)*tan(b)".  As I am a scientific and systems programmer and do *not* work on GUIs, that ability is *very* important.  When I was writing a Kirchoff prestack time migration on Tru64, the thing I missed most was the ability to examine the returns from intrinsic functions. 

Doubtless someone will pop up and claim you can do that with gdb.  All I have to say to that is, show me how.  I'd be delighted.  But my experience with gdb is it's usable where nothing else is.  But it's just barely usable in the best of circumstances.

The single biggest problem in practical computing is terminal complexity.  It is simply not possible for anyone to have a clear understanding of what's  going on.
 

Offline LapTop006

  • Supporter
  • ****
  • Posts: 467
  • Country: au
Re: Gnu/Linux Considered Harmful
« Reply #184 on: November 17, 2018, 03:24:21 pm »
As for "recent directions" and "newfangled crap" Rob Pike (and Ken Thompson) are currently looking after Go at Google...

Nah, Rob's been focused on other stuff for a while now, he certainly still keeps up on it, but it's not his main job.

I work at Google Sydney and often chat with Rob.
 

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23018
  • Country: gb
Re: Gnu/Linux Considered Harmful
« Reply #185 on: November 17, 2018, 03:46:26 pm »
Interesting stuff. Thought Rob was benevolent dictator for life on Go.
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: Gnu/Linux Considered Harmful
« Reply #186 on: November 18, 2018, 10:58:21 am »
It never ceases to amaze me how BSD license proponents say that the GPL license is keeping Linux from being more widely adopted.  The statistics just do not support that argument.  Theory is always nice and beautiful, and opinions interesting, but let's face it: reality overrules wishful thinking.

The real problem is that the amount of code that needs to be usable
This.  The inertia of the existing codebase is absolutely gargantuan.

debugger that will let you do a "print cos(a)*tan(b)".  As I am a scientific and systems programmer and do *not* work on GUIs, that ability is *very* important.
That does also work in gdb (verified on gdb-7.11.1); you can print expressions in the source language. You can also extend gdb with Python functions. For example, (gdb.lookup_symbol('a'))[0].value(gdb.selected_frame()) will evaluate to the Pythonic value of variable a in the current frame.

No, I'm not saying gdb is particularly good debugger -- I don't much like it myself, and prefer a modular/unit verification approach rather than debugging whenever I can --, but I do similar scientific/systems work (mostly in C), and have found this useful on a couple of occasions.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #187 on: November 18, 2018, 12:08:38 pm »
It never ceases to amaze me how BSD license proponents say that the GPL license is keeping Linux from being more widely adopted.  The statistics just do not support that argument.  Theory is always nice and beautiful, and opinions interesting, but let's face it: reality overrules wishful thinking.

The real problem is that the amount of code that needs to be usable
This.  The inertia of the existing codebase is absolutely gargantuan.

debugger that will let you do a "print cos(a)*tan(b)".  As I am a scientific and systems programmer and do *not* work on GUIs, that ability is *very* important.
That does also work in gdb (verified on gdb-7.11.1); you can print expressions in the source language. You can also extend gdb with Python functions. For example, (gdb.lookup_symbol('a'))[0].value(gdb.selected_frame()) will evaluate to the Pythonic value of variable a in the current frame.

No, I'm not saying gdb is particularly good debugger -- I don't much like it myself, and prefer a modular/unit verification approach rather than debugging whenever I can --, but I do similar scientific/systems work (mostly in C), and have found this useful on a couple of occasions.

gdb has always printed expressions, but not if printing the expression required intrinsic function calls.  Did you try my example or a different expression?  I'd be *very* happy if it did as the Forte/Sun/Oracle development suite doesn't look to have a rosy future,

I mostly worked in C and it is my language of choice until I have to do complex arithmetic at which point I shift to F77 as the rules are cleaner and simpler.  And multidimensional arrays in C are awful.  I'd guess that the FORTRAN codes I worked on have finally been retired, mainly because of mergers in the oil industry.  But I had a 125,000 line mapping application that had particularly good logic for avoiding overlaying well names around a platform.  The original author tried to rewrite it in C or C++, but never got it right.  So they kept using the old version.   Because of a bug in the side label logic and the size of the map to be printed, it was splitting the plot into two sheets with 1-2" of the map on a 40+" sheet of paper. IIRC it took me a week to sort that out.  And it was pretty good code compared to the stuff research scientists write. 

The hard part of numerical codes is the indexing in many dimensions of matrices which in seismic processing is pretty much everything.
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: Gnu/Linux Considered Harmful
« Reply #188 on: November 18, 2018, 01:24:58 pm »
Did you try my example or a different expression?
The exact example; I verified it with a tiny test C program with a and b as variables.  If the binary defines say double f(double a, double b), you can do print f(1.0, 2.0) and in general, use f() in the expressions.

And multidimensional arrays in C are awful.
Fully agreed. They're horribly implemented in existing libraries (BLAS/LAPACK/ATLAS, GSL) also.

And it was pretty good code compared to the stuff research scientists write.
Fortran95 and later is a pretty good tool for research scientists, because straightforward code generates pretty efficient binaries, and it is easy to write somewhat readable code.  Other than that, the code research scientists have written that I've worked on, is pretty stress-inducing, if you care about robustness, efficiency, and maintainability.

The hard part of numerical codes is the indexing in many dimensions of matrices which in seismic processing is pretty much everything.
I mostly deal with two-dimensional matrices, with submatrices or vectors as views into them.  (I've shown some data structures elsewhere that show how to make "views" first-class citizens, and still allow efficient computation on current CPUs; boils down to using (row*rowstride + col*colstride) for addressing, and a separate reference-counted data structure to own the actual numerical data each matrix refers to.) And also with huge distributed sets of 3D particle data, which requires all sorts of ordering tricks to have any hope of cache locality; the datasets are so large that without cache locality, memory bandwidth is stricter bottleneck than the interparticle computations (various potential models).

Current simulators like lammps and gromacs also do not interleave computation and data transfer, so they waste quite a lot of wall clock time waiting for data to transfer.  I have a better core structure that avoids that, but the inertia in materials physics is too big to get any kind of support.  It is easier to get money for hardware, than for software development, and that shows.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #189 on: November 18, 2018, 06:14:40 pm »

That's really good to know as I'm mostly going to be focused on ARM code for the next year or two working on FOSS DSO FW.  So using gdb is unavoidable.

A friend uses f95 a lot and really likes it.  I never found a suitable book, but he sent me the last draft of the standard, so that might actually be useful.  I want a compiler writer's description.  But truth is I'm more likely to be fixing old computational electromagnetics codes than writing anything new.

Research scientists almost universally can't program.  I once had a simple subroutine, no comments but author and date, that I had to go get the original paper to figure out if it was working in the time domain or the frequency domain.  It was a project where the "spec" was "what the old program does".  The IT folks were under the delusion at the start that they would just write a new UI and use the old research code.  I very quickly disabused them of that idea.  The old research code was most of the cause of the problems with the existing system.

When I started seismic was 2D.  Now it's all 3 & 4D and even more if it's multicomponent data and you have 3 time series for each sensor.  Then you're looking at 9-12D matrices.

Cache architecture is a real nightmare to deal with.  It's painfully easy to cause a cache miss on *every* data access which then results in 8-16x more traffic to main memory.   Unfortunately, it's also *very* hard to find a viable algorithm for something like a Kirchoff migration which effectively sums 125,000 points on the surface of a hyperbola of rotaion into each output sample.

I've certainly written plenty of C to access portions of FORTRAN arrays.  But aside from being clunky, it's painfully easy to make an "off by one error".   As a summer intern at the ARCO Plano Research Center I was asked to help out with a 3D FFT running on an Intel i386 Hypercube.  I told the scientist to feed it two inputs, a cube of ones and a cube of zeros with a one in the center.  He found the error in a matter of minutes when he ran those two cases.

He was gone by the time I got there, but there was a scientific programmer who had worked there who was legendary for using "i, ii, iii, j jj jjj etc" for variable names naturally with no comments and implicit typing.  I hope that whoever came up with the idea of implicit typing was consigned to the lowest level of hell.

Designing test cases is the hardest thing to do with scientific codes.  You've got to find a problem for which you know the answer analytically.  And that can really require a lot of thought.

Materials physics must be fun.  I got heavily involved in the equations of state for water and hydrocarbon mixtures and the continuum mechanics of porous media.  Someone once wrote a paper about the rise of water in a well whenever a train stopped at the station.  And there are lots of plots of reservoir pressure varying with tidal effects.
 

Offline borjam

  • Supporter
  • ****
  • Posts: 908
  • Country: es
  • EA2EKH
Re: Gnu/Linux Considered Harmful
« Reply #190 on: November 18, 2018, 07:26:06 pm »
It never ceases to amaze me how BSD license proponents say that the GPL license is keeping Linux from being more widely adopted.  The statistics just do not support that argument.  Theory is always nice and beautiful, and opinions interesting, but let's face it: reality overrules wishful thinking.
I never said that. But the obsession with licenses has prevented the Linux world from adopting nice software and often they have poorly reinvented stuff because of that.

The funny thing is, it became something even religious. I would show FreeBSD to the typical young Linux user and the first question would be: "Nice, but what's the license?"

 
The following users thanked this post: techman-001

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #191 on: November 18, 2018, 10:26:08 pm »
A sensible person would think that a program such as mkfs.ubifs with a number of device specific options would have documentation explaining what the command line options were and what information needed to accompany them.  I just spent about an hour trying to find some documentation.  Nothing.  No man page.  A mostly useless slide deck ("ugly because the animation doesn't work").  It looked fine to me except that it did not tell me about mkfs.ubifs.  A Sensible Unix user would expect a man page. Or in the slightly perverse world of RMS, an info file.

But noooo children.  We don't need no stinking man pages or info.  Just use "--help".  Note, I did not read that anywhere.  I poured myself a stiff Scotch and water and tried it in desperation.

I am so sick of running multiple different Linux instances because almost nothing is actually portable.  It's actually *worse* than the workstation wars.  There is so much wrong or out of date crap on the web that it's approaching having negative value.  But it has become the only option because we have a generation or so which has no concept of  *edited* writing in which a knowledgeable person reads the work and identifies errors and  omissions.

Minix lacked virtual memory.  Fred van Kempen implemented it, but Prentice-Hall forced him to take it down because they owned the copyright on the Minix source.   Linus created Linux so it would have virtual memory, but now we have all the baggage but the swap file is only slightly larger than physical memory in a standard install.  My normal Unix install was swap = 4x core.  That allowed me to keep lots of processes in the process table, with all the initialization and history.  If I didn't use them, they'd get swapped out to disk.  No impact other than they were a little slow to start back if I hadn't run them for a while.

The Marx brothers couldn't top this.
 

Offline vtwin@cox.net

  • Regular Contributor
  • *
  • Posts: 175
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #192 on: November 18, 2018, 11:38:01 pm »
There is so much wrong or out of date crap on the web that it's approaching having negative value.

I've made this same exact comment to my boss on numerous occasions. Not long ago I had an issue w/ Exchange Server throwing a bizarre error message. I needed the company credit card to call MS to pay for per-incident support. My boss asked "why? just google it." Um, yeah, why didn't I think of that? Probably because I *did*, and the only postings I could find dated back a decade ago for Exchange 2007 (and we're on 2016)

It's gotten to the point now where I have to sort results by date, and throw out anything more than, say, 3-4 years old. There's crap dating back 15 years in some instances.
A hollow voice says 'PLUGH'.
 

Offline rhbTopic starter

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: Gnu/Linux Considered Harmful
« Reply #193 on: November 19, 2018, 01:49:17 am »
It would be wonderful it that was the only problem.  But most of the time no mention is made of distribution, version, hardware or any of the other relevant factors.  So even if it was posted last week, it's likely to be wrong for your problem.

At my last contract with a super major oil company they were using RH EWS because that's what suite of $100K/seat applications supported.  I had a Sun Blade 1500 for my work for which I wrote all the code.  I was told it was the *only* Sun workstation still in use in the entire company.

My business partner at the time was using the standard issue workstation and PC.  I was quite wide eyed when they had to reboot his system because the display froze.  I'd been having the same problem at home with a machine I'd built.  I'd assumed that is was flaky hardware on my system.  But after I saw a top end IBM workstation do the same thing I bought an Ultra 20 with 3 years of gold level support.  I would have greatly preferred a SPARC system, but the price differential was just too large. Even at gold level, they couldn't resolve everything, but it was very nice. Until Larry took over. Scott McNeally built Sun, but then he destroyed it with ridiculous change of control clauses for senior management.  I think Solaris would be going great guns if IBM had acquired Sun instead of Oracle.

I'll use Linux is I have to, but *only* if I have to.  Windows 7 Pro is a joy relative to CentOS and Ubuntu.  Happily, Debian 9.3 is proving pretty tolerable.

But a friend of mine likes to say  I'm 3 standard deviations out from the norm.  I expect my systems to run for indefinitelyr without being rebooted. An admin at work once asked if he could take my system down to install patches.  It had been up for 467 days.
 

Offline BradC

  • Super Contributor
  • ***
  • Posts: 2104
  • Country: au
Re: Gnu/Linux Considered Harmful
« Reply #194 on: November 19, 2018, 02:17:05 am »
It had been up for 467 days.
Code: [Select]
brad@bkcuk001:~$ uname -a ; uptime
Linux bkcuk001 3.16-0.bpo.3-amd64 #1 SMP Debian 3.16.5-1~bpo70+1 (2014-11-02) x86_64 GNU/Linux
 02:16:32 up 1469 days, 46 min,  1 user,  load average: 0.00, 0.01, 0.05
 
The following users thanked this post: tooki

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: Gnu/Linux Considered Harmful
« Reply #195 on: November 19, 2018, 05:29:56 am »
The funny thing is, it became something even religious.
You ascribe it to religion, because you dislike it for personal reasons.  That is intellectually dishonest.

I explained here the reasons why I myself, and a majority of those who I've talked with that deliberately choose licenses for their own projects and prefer GPL over BSD, do so.  It has nothing to do with religion or following an authority, and everything to do with what one wants the "rules" to be.

The fact of the matter is that BSD allows closed-source derivatives, GPL does not.  If we look at long-term software projects, we find that using BSD/MIT-style licenses helps with initial adoption, while the GPL license attracts more developers in the long term.  It is not religion, it is preference for one set of rules over another.

At this point, every BSD proponent who considers GPL "viral" or "less free" that I've talked to, has started gushing how one set of rules is clearly better than another.  To those arguments, I can only say that reality overrules theory and wishful thinking.  It is one thing to have a philosophical discussion about licenses, and a completely different thing when one discusses which license a particular project should use.


I am so sick of running multiple different Linux instances because almost nothing is actually portable.
Unfortunately, at the same time, that is exactly why Linux is so popular on servers and scientific computing: because it can be customized to suit the task at hand, and not the other way around.  It's just that it is also abused, in a way: choices are made, but never documented.

One detail that consistently trips those used to Unix configuration or Linux Standards Base (when it was still a thing), is that a lot of configuration is under /usr or /var, rather than /etc, when systemd (or any of the services now subsumed into systemd) is used.

It is not difficult to create portable programs for Linux, but it is more work.  Very few bother.

I claim the underlying problems are caused by developer shortsightedness, and lack of care for the long term.
 
The following users thanked this post: Karel, Masa

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23018
  • Country: gb
Re: Gnu/Linux Considered Harmful
« Reply #196 on: November 19, 2018, 08:13:07 am »
Borjam is correct though. Most of the maintainers I’ve spoken to over the years picked GPL because it was the one they had heard about. Not because of explicit ideology. If that’s not a faith argument I don’t know what is.

The reasons the closed source derivatives exist is because the maintainers of the actual projects usually do a piss poor job of maintaining them. This hasn’t traditionally been visible until GitHub came along and you can see the acres long lists of tickets with critical problems even on some major projects. They don’t respond to users, don’t accept patches because they didn’t write them themselves and generally treat users like dirt. This means you have to fork and fix occasionally. Even for trivial things. Then brow beat the maintainers to accept your patch (usually with the hordes of other crying and bug cancer ridden users). Either that or maintain a public fork and publish changes forever. This is a massive problem which actually usually at this point discounts a product from being used in a lot of corporates because the overheads of a tiny decision to import one little GPL licenses component can cost more than an oracle license over a few years.

BSD? Patch in situ, carry on concurrently while your patch is sitting there for a decade. Or if you want to make a private extension that benefits no one other than yourself or is critical to your business then you can without having to deal with linking bollocks.

Personally if I write something I want people to use it and do what they want with it. I will accept patches. I will fix things. I will respond to users. If someone wants to make a better version with their vision, knock yourself out. If someone wants to include it in some other software, knock yourself out too.

Just attribute it to me in some way and have fun. BSD for me. Isn’t that really what we want?

No chess, no egos, no ideology, no debates on the freeness of beer. Solving problems that is all.
« Last Edit: November 19, 2018, 08:15:03 am by bd139 »
 

Offline borjam

  • Supporter
  • ****
  • Posts: 908
  • Country: es
  • EA2EKH
Re: Gnu/Linux Considered Harmful
« Reply #197 on: November 19, 2018, 08:46:45 am »
The funny thing is, it became something even religious.
You ascribe it to religion, because you dislike it for personal reasons.  That is intellectually dishonest.
Preposterous. Where did I say that I dislike the GPL license?

I dislike the Linux ecosystem and what it has brought to software development because of its own technical demerits, which is very different.

And none of the people I have in mind were developers. They were users. Of course it you are developing something and releasing it to the world you will likely choose a license. But ¿just using the software?

I think that the Free Software Foundation has done a lot of good. However, should all software be GPL licensed? Not at all.

And I didn't choose BSD because of the license. That's irrelevant.
 

Offline bd139

  • Super Contributor
  • ***
  • Posts: 23018
  • Country: gb
Re: Gnu/Linux Considered Harmful
« Reply #198 on: November 19, 2018, 09:16:21 am »
I think no software should be GPL licensed as it removes practical freedoms. Call me religious if you want everyone :)
 

Offline Nominal Animal

  • Super Contributor
  • ***
  • Posts: 6173
  • Country: fi
    • My home page and email address
Re: Gnu/Linux Considered Harmful
« Reply #199 on: November 19, 2018, 10:55:01 am »
Borjam is correct though. Most of the maintainers I’ve spoken to over the years picked GPL because it was the one they had heard about. Not because of explicit ideology. If that’s not a faith argument I don’t know what is.
We must run in different circles, because that is completely different to what the people I've talked about this have said.  Cultural differences?

The reasons the closed source derivatives exist is because the maintainers of the actual projects usually do a piss poor job of maintaining them.
Bullshit.  Closed source derivatives exist if and only if someone believes they can make money off of them. (There is nothing wrong with that, in my opinion; just let's be honest here.)

This has occurred time and time again, with both operating systems (BSDs), and larger projects like the Apache web server.  The history of the closed-source Apache derivatives is especially funny: they just couldn't keep up with the open source upstream.

I believe it is also at the core why the BSD variants suffer from a lack of developers: a company like Apple, having their own derivative, can poach the best developers, and have them work on their derivative only.  At least when companies hire Linux kernel devs (and they do, often), those devs still keep contributing to the mainline kernel.

[...] BSD for me. Isn’t that really what we want?
Don't change the subject! It is a completely different thing to discuss what licenses one wants to use for their own projects, than to discuss what licenses others should use, and why.

(I really don't mind what license or rules people choose for their projects; I (and those I've talked to) evaluate each one on a case-by-case basis, with the license requirements part of the "cost".  Like I already mentioned, I personally use several different licenses for my own work, depending which set of rules I believe works best for each case.  I'm very pragmatic that way.  But this is not the topic at hand.)


The funny thing is, it became something even religious.
You ascribe it to religion, because you dislike it for personal reasons.  That is intellectually dishonest.
Preposterous. Where did I say that I dislike the GPL license?
That's how your text reads to me, that's all.  If you want, I can reword myself:

"You seem to find it distasteful and anti-intellectual when people choose the GPL license, because you ascribe the choice to religion."

There is nothing religious about the license choice.  Even if people were to simply pick the license they've heard most about, that would be lazy and conformist, perhaps; nothing at all to do with religion.  And you know that; yet you use the term as a denigrating label.  That is, as far as I understand, the very definition of intellectual dishonesty.  Low manipulation using terms loaded with emotional connotations.

I don't play those social games myself.  I prefer direct, blunt, honest communication.

However, should all software be GPL licensed? Not at all.
Why are you telling that to me? I've already stated I use several licenses myself, so obviously I don't think all software should be GPL licensed.

BSD license is an excellent choice for example for low-level interface libraries, language standard libraries, device drivers, and so on, where the initial adoption is more important than the risk of having incompatible closed-source derivatives.

Being open source at all is not always an option, either.  Sometimes selling licenses is the only way to fund the development and support.  It is a perfectly valid business model, nothing wrong in it.

If the Linux ecosystem used BSD instead of GPL, it would not be where it is right now.  Claiming otherwise is ignoring FreeBSD, OpenBSD, and the other BSD variants, and their project history; wishful thinking not based on real world experiences.  That is why I find this kind of discussion so hilarious and annoying at the same time.

It is also completely fine to dislike the Linux ecosystem, for whatever reason.  But to claim that it would have worked out better if they had chosen a different license, or something along those lines, is just unfounded daydreaming out loud.  Hard statistics already show that choosing GPL over BSD for the base license for an OS has lead to a larger development community, and better results (as in no BSDs in top-500 supercomputers, for example).


I think no software should be GPL licensed as it removes practical freedoms. Call me religious if you want everyone :)
No, but idealistic, maybe.  You think everything would be better if licensed under BSD or similar license, but you have no basis for that; only hope.  (If you said it was a truth revealed to you in a vision or something like that, then I'd call you a religious nutter.)

Would we have GCC and the other GNU tools without the GPL?  It is not clear at all.  It is silly to assume that you could change the license at the base of a community, with nothing else changing.  Humans don't work like that.

The concept of fairness seems to be quite built-in to hominids, apes, and monkeys:


I suspect (but have not gathered any real facts behind that suspicion) that a lot of developers who prefer GPL over BSD do so because they perceive GPL as being fairer, in the sharing sense (i.e., that GPL enforces symmetric sharing, whereas BSD does not).  If true, it could explain why GPL-based open source communities seem to attract more senior developers than BSD-based ones do.  (The commercial closed-source software world works under completely different rules, of course.)

(Popularity being something completely different again.  I, for example, am not at all interested in making Linux more popular.  All I want is better tools, and specifically tools that I can mold to my own workflow, and don't demand I conform to the tools.  If it means Linux will never be popular on the desktop, fine by me: I don't care if other people draw with fingerpaints and eat glue, as long as I don't have to.)
« Last Edit: November 19, 2018, 10:58:25 am by Nominal Animal »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf