undefined | Better HN

0 pointsabetusk1y ago0 comments

1997-1998 was about the time I first installed Linux (slackware) from a stack of 3.5" floppy disks. By then, Linux had picked up enough momentum, which is why, I guess, you and I both had access to CD/floppy installation methods.

The folklore around the Linux/Minix debate, for me, was that "working code wins" and either microkernel wasn't as beneficial as was claimed or through grit and elbow grease, Linux pushed through to viability. But now I wonder how true that narrative is.

Could it have been that FOSS provided the boost in network effects for Linux that compounded its popularity and helped it soar past Minix? Was Minix hampered by Tanenbaum gatekeeping the software and refusal to cede copyright control?

To me, the licensing seems pretty important as, even if the boost to adoption was small, it could have compounding network effects that helped with popularity. I just never heard this argument before so I wonder how true it is, if at all.

0 comments

3 comments · 1 top-level

lproven1y ago· 2 in thread

> The folklore around the Linux/Minix debate, for me, was that "working code wins" and either microkernel wasn't as beneficial as was claimed

Hang on. That does not work.

You need to be careful about the timeline here.

Linus worked with and built the very early versions of the Linux kernel on Minix 1.

Minix 1 was not a microkernel, only directly supported 8088 and 8086 (and other architectures, but the point here is not 80286 or 80386, so no hardware memory management) and it was not FOSS.

Minix 2 arrived in 1997, was FOSS, and supported the 80386, i.e. x86-32.

Minix 3 was the first microkernel version and was not released until 2005.

You are comparing original early-era Linux with a totally different version of Minix that didn't exist yet and wouldn't for well over a decade.

In the early 1990s, the comparison was:

Minix 1: source available, but not redistributable; 16-bit only, max 1MB of RAM, no hardware memory protection, and very limited.

Linux 0.x to 1.x: FOSS, 32-bit, fully exploited 32-bit PCs. 4GB of RAM if you could afford it, but could use 4MB - 8MB for normal non-millionaire people.

abetuskOP1y ago

I'm pretty ignorant of the timeline and I'm only going off of some vague memory and what Wikipedia says.

Note that the OP is from 1992, so Tanenbaum was arguing for micro-kernels well before the Minix 2 1997 release.

From Wikipedia's entry on Minix 1.5 [0]:

> MINIX 1.5, released in 1991, ... . There were also unofficial ports to Intel 386 PC compatibles (in 32-bit protected mode), ...

I found an article online that dates from 1999 but referencing a comp.os.minix post from Tanenbaum from 1992 where Tanenbaum clearly states MINIX is a microkernel system [1]:

> MINIX is a microkernel-based system.

Further, I don't see any reference of Minix 2 being released as FOSS in 1997. Wikipedia claims Minix 2.0.3 released in May 2001 was the first version of MINIX released under a BSD-3 license [0]:

> Version 2.0.3 was released in May 2001. It was the first version after MINIX had been relicensed under the BSD-3-Clause license, which was retroactively applied to all previous versions.

From Wikipedia's entry on "History of Linux" [2]:

> In 1991, ..., Linus Torvalds began a project that later became the Linux kernel. ... . Development was done on MINIX using the GNU C Compiler.

> On 25 August 1991, he [Linus torvalds] ... announced ... in another posting to the comp.os.minix newsgroup

> PS. Yes - it's free of any minix code, ...

So, I don't really know if Tanenbaum was talking "in theory" about where best to allocate effort and/or if Minix 1/2 were actually a microkernel design, but it seems so. I'm also pretty ignorant of whether Minix 1/2 could be used on 80286 or 80386 chips.

Though I'm very fuzzy on the details it does seem like the sentiment remains. It looks like Torvalds work on Linux was, either directly or in large part, due to the restrictive licensing of Minix [3]:

> Frustrated by the licensing of Minix, which at the time limited it to educational use only, he began to work on his operating system kernel, which eventually became the Linux kernel.

In a 1997 interview, Torvalds says:

> Making Linux GPL'd was definitely the best thing I ever did.

[0] https://en.wikipedia.org/wiki/Minix#MINIX_1.5

[1] https://www.oreilly.com/openbook/opensources/book/appa.html

[2] https://en.wikipedia.org/wiki/History_of_Linux

[3] https://en.wikipedia.org/wiki/Linux#Creation

[4] https://web.archive.org/web/20070826212454/http://www.tlug.j...

lproven1y ago

> I'm pretty ignorant of the timeline and I'm only going off of some vague memory and what Wikipedia says.

I can see that. That's what I was addressing.

This was not a technical contest. It was not about tech merit. It was not about rivalry between competing systems. They were not competing systems. They never have been and still are not.

When both Linux and Minix were options on the same hardware, Linux was FOSS and capable but incomplete, Minix was not FOSS, not capable, but did work and was available, just not redistributable.

It was not a competition.

AST was talking about the ideas and goals of OS research when he wrote.

Then, later, in subsequent versions, he first made Minix really FOSS (ignoring quibbling about second-decimal point versions), and then later, he rewrote the whole thing as a microkernel as a demo that it was possible.

It is arguably more complete and more functional than the GNU HURD, with a tiny fraction of the time or people.

Minix 3 was AST making his point by doing what he had advocated in that thread, about a decade and a half earlier.

j / k navigate · click thread line to collapse