that old GNU/Linux argument

Timothy Murphy gayleard at
Wed Jul 23 00:16:12 UTC 2008

Björn Persson wrote:

> Timothy Murphy wrote:
>> I think it is misleading to say that Torvalds was "dissatisfied" with
>> Minix. The fact is, Tanenbaum (bizarrely) declined to port Minix to the
>> 386 on the absurd grounds that there were millions of 286's around the
>> world, and people would continue using them indefinitely.
> "Port" can't be the right word here. A 386 runs 286 code natively so
> there's no need to port it. You probably mean that he didn't want to make
> use of the more advanced features of the 386.

Possibly "port" is not the right word.
But Tanenbaum explicitly said he would not write a 386 version of Minix
for the reason I gave above.

> What I've heard is that Minix was meant for teaching. Tanenbaum
> deliberately made it simple so that students could understand the whole
> kernel in reasonable time.

That is what Tanenbaum said, yes.
But I think the true reason was that he just got fed up with Minix,
and wanted to get on with other things.

> Those who wanted Unix on their PCs used Minix because it was the only
> Unix-like OS available for PCs. Some people wrote patches to improve it,
> but Tanenbaum didn't incorporate their patches because he wanted to keep
> Minix simple. Tanenbaum's license wouldn't allow anyone to redistribute
> patched versions of Minix, so those who wanted an improved Minix had to
> get the patches separately and apply them.
> So I don't know if Linus was dissatisfied but some people certainly were.
> There was a pent-up demand for a more advanced Unix-like OS for PCs that
> would allow people to help improving it, and then Linux appeared on the
> scene and provided the last piece that made GNU usable.

What I should have said is that Linus was not dissatisfied with Minix
in the sense that he didn't think it was very good.
He was dissatisfied - and so were lots of other people, including me -
that Tanenbaum refused to continue development in a rational way,
by using the facilities available in the 386.

He could easily have done this, and if had done
we would all be using Minix today instead of Linux.

In my opinion, Minix code was far better than Linux code.
But Tanenbaum lacked the "common touch" that allowed Torvalds
to get a large number of clever people (like Alan Cox)
to write code _with_ him rather than for him.
Tanenbaum's setup was very different, with the Master (himself)
assigning coding tasks to his students.

More information about the fedora-list mailing list