[TUHS] Why Linux not another PC/UNIX [was Mach for i386 ...]

Larry McVoy lm at mcvoy.com
Thu Feb 23 09:13:20 AEST 2017


On Wed, Feb 22, 2017 at 05:56:38PM -0500, Clem Cole wrote:
> BTW: you pick on BFFS (aka UFS).   I was under the impression that pretty
> much that was an entire rewrite.  

Nope, that's my wheelhouse, I spent a lot of time in there.  I rewrote
bmap() for the extent based file system work:

http://www.mcvoy.com/lm/bitmover/lm/papers/SunOS.ufs_clustering.pdf

so I know that code very well.  The bmap() from research is the bmap()
in BSD.  It's bit for bit, including comments, identical.  There were
others but I sort of lost interest at that point, it was stolen code
in my opinion.

> code Berkeley had written during the joy's speed up work (even things like
> bmap).   I did not think that came from 32/V.   The key is the changes were
> done a little a time.  Some in BSD, some more in BSD 2.0, more in 3.0 ....
> by the time of 4.2 the feeling was most of the kernel was different from
> what Ken and Dennis had originally written.

A lot of it was different.  But enough of it was the same that it was 
stolen code in my opinion.  And I don't say that lightly, I admire and
respect the BSD guys from CSRG pretty deeply.

> > I'd argue that what you'd get from any reasonable programmer is the
> > naive initial version that works in theory but fails in practice.
> >
> ???Fair enough... and you might be able to make a reasonable living as a
> expert witness ;-)

I took early retirement, I spend a lot of my time on tractors these days:

http://www.mcvoy.com/lm/tractor-tree.jpg

If someone offers me a pile of money to be an expert witness, I suppose.
But quite frankly, that sounds like being an adult.  Not a lot of fun
in my opinion.

> > In my mind, the BSD guys cheated.
> 
> ???I understand and I see your point, although here is where we disagree.???  I
> lived it differently, since I was part of it.  I think it was like the wolf
> that became the dog. A very slow process.  But at some point, there was a
> change and the wolf stopped being a wolf and started being a new thing, a
> dog.

I'd so love to agree with you.  But the code diffs I did do not support
that view, there was more than enough unchanged to call it cheating.
I think that if I spent 20 minutes with you diffing code you would agree.

> ???Fair enough.  To you its still a wolf and I can see that and if I did see
> it as a wolf, I suspect I might agree.  But I think having lived so much
> working being done outside of AT&T that was not getting credit and that
> "DNA" was being "stollen" by AT&T in my mind, it worked both ways.

So there we can agree.  In no way do I hold AT&T blameless, if you want
me to rail on them I can do that.   Kind of an easy target, they really
didn't understand what they had or what to do with it.  It was better
when it was Bell Labs and Bell Labs had a charter (didn't it have to 
do with being a monoply) that said they released their patents for
free (or something like that, right?).  When the suits came in it all
got screwed up.

And for the record, BSD moved the state of the art forward, I know that.
I'd much rather be stuck on 4.3 BSD than SVr3 (and even SVr4).  They redid
a lot and added sockets and at least imagined mmap().  A BSD kernel
was a nicer place to be than a AT&T kernel.  So I mean no disrespect
to their work.  I just don't think they rewrote as much as they claimed
they did, it was not a total rewrite, there was more than enough still
there that AT&T could have prevailed with a copyright case.

--lm


More information about the TUHS mailing list