Confessions of a Recovering NetBSD Zealot

debilo writes, "ONLamp.com is featuring a lengthy interview with Charles M. Hannum, to Slashdotters probably best known for his wake-up call aptly titled The Future of NetBSD that generated a rather vocal discussion. In the interview, Charles speaks about his role in and the beginning of The NetBSD Project, shares his thoughts on software licenses, discusses the popularity of Linux and its development model, and further addresses the problems that NetBSD is facing. Some notable quotes include: 'If I were doing it again, I might very well switch to the LGPL. I'll just note that it didn't exist at the time.' And: 'There was a lot of FUD around this issue — some of it from Linus, actually — and it did cause us some problems.'"

Comments

Re:We need a NetBSD

[sudog]

Base install for NetBSD is about 23MB, or 47MB with development tools, 55MB with manpages, and 112MB for all the above plus X.

I can get it installed on older hardware in less than 5 minutes, including the boot time for the floppies. I can get it installed on modern Opteron-based badass hardware in about half that. That's pretty cool.

And you're being very short-sighted about other architectures.

GPLv3 OS

[r00t]

You can't convert Linux. GPLv2 and GPLv3 both prohibit extra restrictions.

You could create a GPLv3 fork of NetBSD though. That might revive NetBSD. You might just take the kernel though, letting distributions form around it. Debian already supports Hurd and FreeBSD kernels; they could do a NetBSD one as well.

Of course you'd need to find a name other than "NetBSD".

Ideas: NetOS, NotBSD, Netix, Netrix, Netux, Nettle, WebBSD...

The answer is simple...

[Lost Found]

...but you may not like it. The replacement is and will continue to be Linux, which is already more portable than NetBSD, has far greater mindshare, performance, scalability and functionality.

X11 is heavyweight?

[steveha]

X-Windows needs to be replaced with something more light-weight (i.e. single-user with direct access to the multimedia hardware).

Really? Can you please point me to some numbers that demonstrate this point?

X11 was invented in the bad old days, running on UNIX systems less powerful than today's PDAs. As I understand it, it's actually quite lightweight. Certainly the network transparency features don't cost much, because when you run the X server and the X client software on the same computer, they communicate by using domain sockets (which are very lightweight). Both Microsoft Windows and Apple OS X have abstraction layers that isolate the graphics hardware; do you have some numbers showing that X11 has significantly more overhead than those abstraction layers?

The latest versions coming out of X.org now have support for features similar to what OS X does: applications are rendered into offscreen buffers, and the buffers are composited together (with transparency effects, or other special effects if you desire). So, X11 is no barrier to cool eye-candy either.

The worst thing about X11 used to be way it was managed (under Xfree86). Now that the project has moved to X.org and has been revamped, progress has sped up a lot.

steveha

Re:X11 is heavyweight?

[TheRaven64]

Certainly the network transparency features don't cost much, because when you run the X server and the X client software on the same computer, they communicate by using domain sockets (which are very lightweight).

Not that lightweight. Fortunately, in the '80s, MIT released the shared memory extension which took away most of that overhead and has been standard in X servers for over a decade. The problem with X11's network transparency is that it is at the wrong layer. X11 puts network transparency between the view and the frame buffer, when it should be between the view and the controller. Sun realised this with NeWS, but did a typical Sun and said 'Hey, we've got this great technology! How can we market it in such a way that it never goes anywhere?'

Architecturally, there are quite a few things wrong with X11. The easiest solution is the one that Apple took; throw it away and replace it with something new. That isn't really a good idea for *NIX, however, since there is a lot of legacy software that uses X11. Fortunately, Keith Packard seems well aware of the shortcomings of X11 and has a set of incremental improvements that address them.

Re:What is really needed

[bfields]

If Linus continues to dig in and refuses to accept GPLV3 with its anti-DRM provisions

Good grief:

1. GPLv3 isn't even finished yet. This is a little premature.
2. Linus states his opinions very clearly and forcefully. That *doesn't* mean he isn't willing to change his mind. He's done it before. BUT:
3. Even if Linus did "accept GPLV3", he doesn't own the copyright in most of the kernel himself at this point--for each piece of the code he doesn't own copyright in, somebody would have to either track down the copyright-holder (or their heirs, if necessary...) and get their permission, or rewrite their code. The kernel is really big; the chances of this happening are extremely small.

Re:All this comes as I think about going to BSD

[Nimrangul]

DesktopBSD and PC-BSD are not forks of FreeBSD, they are add-ons to FreeBSD releases, they are prepackaged versions of FreeBSD, similar to your Linux, "distros," mix KDE with random home-brewed tools and pour into release version 5, 6, or whatever of FreeBSD, serve. DragonFly BSD, while interesting, is not really the system one should be looking into for regular day-to-day use, major design alterations are still underway in the system and as Dillon has said, it's not going to really be ready for a while. NetBSD as you've recently noticed has some issues, it's had stability problems on it's vaunted umpteen jillion platforms for going on the better half of a decade and at best users and developers have been ignoring much of it, the developers focus on making things cross-compile rather than making sure they natively compile and actually run.

If you want to try a BSD, FreeBSD is the most Linux distribution-like of the BSDs, while OpenBSD is the most BSD-like of the BSDs, it's a matter of if you like the old Unix stuff or the newer Linux stuff. Generally speaking FreeBSD is the less GNU-style free, as in that whole freedom schtick, of the two, with better performance and more bells and whistles, while OpenBSD is the more secure, stable, conservative of the two. OpenBSD has better overall documentation, while FreeBSD has several really nice books, like the DAIOTFOS and the Handbook.

Re:What is really needed

[SEE]

Actually the linux kernel could be forked from the existing code base licenced as GPLV2 with ongoing contributions to the new kernel licenced as GPLV3.

Actually, no.

The Linux kernel is specifically licensed under GPL v2 only. Due to the deliberate design of the copyleft in the version 2 GPL, it would be illegal to distribute that code in a kernel where portions of the kernel could not be redistributed under the terms of the GPL v2. The inability of GPL v2 code to be put under a more restrictive license without permission of the copyright holder is a deliberate, designed-in feature, and no FSF backdoor exists.

That means, all the "GPL v3" code in your "forked" kernel would either have to include specific permission to be used under the GPL v2, or none of it could be the current GPL v2 Linux kernel code. In the first case, then, anybody could grab any of the code from your "fork" and distribute it under the v2 only, defeating the whole point of forking a project to add v3 DRM restrictions. In the second, you wouldn't be forking Linux, you'd just be writing a new kernel.

You want a GPL v3 GNU kernel? You can start with HURD, or with one of the nonproprietary BSDs (4.4 Lite, Free, Net, Open), or a few more obscure things. But you can't start with Linux if Linus says no, because the FSF set up the GPL v2 that way.

Re:So what is the next Toster OS?

[hubertf]

sorry to bust your bubble, but have you heared anyone actually working on NetBSD say it's dying? It's thriving more than ever, and just a few people that can't get things their way propose its death. If those few people could get over their desire to get into "politics", this whole discussion would not exist.

NetBSD does live, it will live,
but it also needs people to do the work, not just talk.
Join in!

  - Hubert

Re:We need a NetBSD

[mdhoover]

The whole point of NetBSD is portability. If it weren't for portability, NetBSD might as well not exist. But the problem is Linux has taken over as the portability leader and has a huge margin.

Yes and no. Most people who make this comment should add the caveat "for 2.2.x/2.4.x series kernels with glibc 2.2.x/2.3.x w linuxthreads".

If your hardware is new and relatively common, linux kernel will most likely have a current working port.
Move off the path well travelled and you are in for a lot of strife.

My favorite 2 examples to bring up, mainly due to the massive number of units still in existence are

1. mac-m68k - last worked kernel 2.2 series, broken 2.4 and 2.6 (ADB broken, serial console broken)
2. sparc32 sun4m - currently SMP support broken in 2.6 kernels for supersparc, hypersparc just broken.

At least sun4m looks like it will get fixed, dont hold your breath for the m68k macs.

Dont get me wrong, I am a linux zealot, but linux' rate of change pretty much ensures that older architectures will drop out of the tree due to lack of maintenance, be it in kernel or glibc.
NetBSD still runs, and you can be pretty assured will continue to run, on most hardware you point it at (including your old vaxen ;-) sure beats VMS ).