[TUHS] A Unix Code Walkthrough by Ken Thompson

[Jonathan Gray]

On Sun, Mar 16, 2025 at 05:26:19PM +1100, Jonathan Gray wrote:
> On Mon, Jan 13, 2025 at 05:23:11PM +1100, Jonathan Gray wrote:
> > On Mon, Jan 13, 2025 at 03:11:21PM +1000, Warren Toomey via TUHS wrote:
> > > On Sun, Dec 29, 2024 at 07:40:15AM +1000, Warren Toomey via TUHS wrote:
> > > > Hi all, I've just received a set of MP3 recordings from Bob Kridle. He says:
> > > > 
> > > >     These are recordings of Ken Thompson doing a read through of one of
> > > >     an early UNIX kernel code listing with a group of grad students at
> > > >     UC Berkeley while he was a visiting prof. there.
> > > > 
> > > > The date is roughly 1975. I've put the recordings here along with his
> > > > e-mails about the recordings:
> > > > 
> > > > https://www.tuhs.org/Archive/Recordings/1975_Unix_Code_Walkthru/
> > > 
> > > Eric Allman has kindly supplied a scan of the source code that ken
> > > provided for the walkthrough. I've added it to the above location.
> > > It's interesting in that there are comments at the start of each
> > > function - the existing 5th Edition code that we have does not
> > > have these comments.
> > 
> > With alloc.c the comments are there in v6 and the copyright line
> > removal matches the changes between v5 and v6.
> > 
> > The code appears to be somewhere between v6 and the "50 changes" tape
> > described by Mike O'Brien in:
> > Applications/Spencer_Tapes/unsw3.tar.gz
> > usr/sys/v6unix/changenotes
> 
> I've retyped the source code to make it easier to compare:
> https://github.com/jonathangray/v6-allman
> 
> Some of the changes are in other v6-derived trees.  Some such as the
> bootstrap and parity changes I couldn't find elsewhere.

A scheduler change from the Naval Postgraduate School, described by Ken as
"this is a fix from Monterey, and I think it's wrong".
Tape 1 (Cleaner) ~ 43:17

Mentioned in the November 1976 UNIX News
https://archive.org/details/unix_news_november-1976/page/n1/mode/2up
tuhs Documentation/Usenix/Early_Newsletters/197611-unix-news-n11.pdf

"Also available by separate request only is a new scheduler, which scans
the process table twice; the first pass throws out a process only if
that process alone will make room for the new process.  NPS claims 25%
improvement on a 48K machine, Ken Thompson thinks otherwise.  You will
have to judge for yourself."

And in February 1976, though the text in the scan is hard to read
https://archive.org/details/unix_news_feb-10-1976/page/n7/mode/2up
tuhs Documentation/Usenix/Early_Newsletters/19760210-unix-news-n3.pdf