More from the archives...

Gene Spafford (
Wed, 17 Oct 90 16:07:49 EST

>From ulysses!smb Mon Oct 26 17:20:45 1987
From: ulysses!smb
Date: Mon, 26 Oct 87 15:42:21 EST
To: cbosgd!mark,
Subject: early netnews

>From moss!rutgers!mcnc!rti!trt Mon Oct 26 15:39:01 1987
Date: Mon, 26 Oct 87 13:52:51 EST
>From: Thomas Truscott <rutgers!mcnc!rti!trt>
To: cbpavo!mark, moss!ulysses!smb
Subject: Early Usenet info

Cyrus Rahman, Duke physics grad student,
recently sent me some early Usenet data
on the disks of the old Duke CS PDP 11/70.
(Physics dept. -- where old computers go to die.)

I am including two interesting files.
The first describes the 'duke' site on Mar 11 1981,
the other is a collection of interesting early net articles.
Amongst the articles is the announcement of decvax.
One interesting thing about those articles is their brevity--
some are only two lines long.
Of course, this was probably a sample of only
the most interesting/useful articles.

Also on the tape was a Usenet Invitation paper,
not the original but fairly old (Mar 1981) .
You have probably seen this excerpt, but in case you haven't:
.IP 1.
Isn't this expensive?
Not at all.
Night time phone costs are about $0.50/3 minutes,
in which time uucp can transfer about 3000 bytes of data (300 baud).
Daily polling then costs $15/Month, which is half what Duke pays
just for an office phone.
.IP 2.
Can Duke really handle all the phone calls?
Sure. We have two call-out lines: at five minutes/call,
we can handle 24 calls/hour.
Other systems can also perform the call-out function.
Remember this when reading the decvax article.

>From swd Wed Mar 11 12:30:45 1981 remote from duke
Here is the current state of the duke system.
contact: Steve Danial (swd), 919-684-3048
CPS Dept, Duke University, Durham, NC 27706
path: we reach ucbvax both through decvax and mhtsa
news: 147 local users of which 24 subscribe to NET news
about: We are a research machine for the Computer Science
Department. Current research includes natural language
processing, VLSI design, and operating systems simulation.
a) Anyone can call us.
We give systems individual logins for accounting,
so we don't publish a uucp name and password.
Interested nodes should contact swd.
b) Typically we will call other sites if they reimburse us
for the phone charges. This is cleared through swd.
uucp connections:
Local nodes:
phs - 11/44 and 11/60
cleps - 11/23
duke34 - 11/34
BTL nodes:
research, vax135, epsvax, topes, chico, mhtsa
unc - U of North Carolina
uok - U of Oklahoma
reed - Reed College
nybca - Ny Blood Center
decvax - DEC

>From jte Sat Apr 26 22:02:50 1980
at-bugs : NET.general
"at" cannot schedule execution on March 1 during leap years.
Fix: />59/s//>=59/

"at" cannot schedule an event during the following year
if requested by month and date.
Fix: Unknown

If the shell script produced by "at" includes the assignment
of exported variables with special characters (e.g. $IFS)
the script can fail.
Fix: A fairly good fix is
/fprintf(file, "%s.n", .ep++);/c
fprintf(file, "%.*s'%s'\n", ix-*ep, *ep, ix);
A better fix would scan *ep to take care of single
qoute marks in a shell variable.

>From jte Mon Apr 28 23:16:57 1980
ac-bugs : NET.general
"ac" as distributed with V7 has several problems.
1) General terminal names are not supported. All terminals
except the console must be named "ttynn". This causes
serious accounting problems as logins on all other terminals
are mapped into the console.
2) Ac looks for "}" to indicate a new date instead of the
correct "{".
3) "Midnight" (-d option) is not always computed correctly.
4) Diagnostic messages are incomplete.

Duke has rewritten ac and fixed these problems.
The new version is available from "duke!~uucp/duke/cmd/ac.c".

>From swd Sun May 4 20:47:16 1980
look bugs : NET.general
look(1) expects /usr/dict/words to be sorted with -d and -f
options. As distributed (on version 7) the wordlist is
not correctly sorted. As a result, "look ba" yields nothing.
Fix: sort -d -f /usr/dict/words -o /usr/dict/words

>From research!dmr Tue May 13 03:27:25 1980
Floating-point bug in V7 Unix: NET.general
J. W. Stevenson of Vrije Universiteit in Amsterdam has
discovered and diagnosed a most interesting bug in
V7 Unix. The problem is that the floating-point registers and
status are sometimes not saved across memory expansion via
sbreak() or stack growth or (less importantly) forks.
The problem only occurs when a swap is needed to
accomplish the expansion.

What happens is that the swap output IO is begun,
and then the registers are saved when the process
calls swtch() from sleep().
It is possible for the image to be written out before
the saving of the registers.

For reasons I do not fully understand situations can
occur in which the bug manifests itself consistently.
It seems to me that the register-saving should be finished
within a few hundred microseconds at most after initiation of IO
while the actual transfer (with ordinary moving-head disks)
will not begin for several milliseconds at least.
Nevertheless we have demonstrated the bug quite consistently;
by the usual sort of coincidence,
one of our machines began showing floating-point problems
the very same week that Stevenson's letter arrived.
(If the controller prefetches a significant amount of
data or if the disk is very fast then the consistency is not

At any rate, as suggested by Stevenson, the problem can be
fixed by adding the following code

if (u.u_fpsaved == 0) {
u.u_fpsaved = 1;

to the following places in sys/slp.c:

before original line 507, "a2 = malloc(coremap, n);", in newproc();

and, after original line 564, "if(a2 == NULL) {", in expand().

>From vax135!ucbvax!mark Tue May 27 23:39:00 1980
uux bug : NET.v7bugs
A bug in uux prevents the - option from working with binary files.
To fix it, in uux.c, search for "fputs". (It's in an if (pipein) {.)
replace the fgets/fputs loop by
while ((c = getc(stdin)) != EOF)
putc(c, fpd);
and declare c somewhere to be an int. (I put the line
register int c;
right after the if (pipein).)

Fixing this bug will allow the new uusend program, which will be released
shortly, to work, allowing the transfer of binary files across indirect
uucp links.

>From unc!smb Thu May 29 19:44:57 1980
at bug : NET.general
The 'at' command insists on creating the spool file in mode 644, regardless
of your umask setting. Thus one is unable to create confidential requests.

>From swd Thu Jun 26 10:58:16 1980
uucp bug : NET.general
In distributed v7 uucp, if the USERFILE grows
to over 15 lines, uucico will die.
Fix: in chkpth.c chg line 14 from
struct userpath Upt[15];
struct userpath Upt[MAXUSERS];
As distributed, MAXUSERS is 20. Chkpth will check
and complain if USERFILE has more lines than MAXUSERS.

>From swd Thu Jul 10 10:42:51 1980
mkconf bug : NET.v7bugs
Some time ago a bug was reported in scanf relating to
scanning the format string "%d%c". (This bug is fixed
in the duke version of stdio.)
It seems that mkconf.c REQUIRES this bug in order to work.
Thus we suggest the following fix to mkconf.c:

diff mkconf.c.good mkconf.c.original
< " trap; br7+10. / 11/70 parity\n",
> " trap; br7+7. / 11/70 parity\n",
< if (n == 0)
< n = sscanf(line, "%s%s%ld", keyw, dev, &num);

>From sid!cf Thu Jul 10 22:33:51 1980
Unix alive on P-E 3240: NET.general
Wollongong Unix has just been installed on a Perkin Elmer 3240
at the University of Melbourne. It is alive and well running
two users. Those tired of VAX delivery times or Unibus Adapter
troubles can take heart.

More info later.

>From vax135!ucbvax!cory:optvax:horton Fri Jul 11 01:11:04 1980
ucbopt lives : NET.general
The optvax at UCB, on the Berknet, has been added to the news network.
This vax is a private vax with a few research projects going on on it.
It is connected off the ucbcory machine, corresponding to the
physical configuration of the Berknet.
Mark Horton

>From trt Wed Aug 13 09:48:33 1980
col : NET.v7bugs
The buffer-flush bug in col(1) that was mentioned in previous news
is due to using an automatic (local) variable for buffering stdout:
char fbuff[BUFSIZ];
should be changed to
static char fbuff[BUFSIZ];
(exit() also works since main() will not return before flushing.)
Col(1) is just one of several V7 programs with this flaw.

In the Duke version of stdio one can bypass the problem with
setbuf(stdout, SYSBUF);
where SYSBUF is a flag indicating buffering is desired.

>From vax135!ucbvax!cory:optvax:jimbo Mon Sep 29 20:40:00 1980
Spice & Splice mail aliases: NET.general
We are creating mail aliases for distribution of occasional messages of
interest and reporting of bugs for both the Spice and Splice circuit
simulation programs. This is not really for user consulting but to
make it easier to report bugs or find out what has been fixed already.
Interested parties who would like to put a name on one or both should
send a note to:

esvax.spice@berkeley (over ARPA-Net)
ucbopt!spice (over "uucp")
esvax.splice@berkeley (over ARPA-Net)
ucbopt!splice (over "uucp")

These aliases should be used for all incoming messages.


>From decvax!shannon Tue Nov 11 17:26:58 1980
DEC on Usenet : NET.general

DEC now has a Unix system on the Usenet uucp network.
Anyone with questions about or problems with DEC hardware
and it's relationship (interaction) with Unix, feel free
to contact us. We are members of the Telecommunications
Industry Group (TIG or Telco) of DEC and are actively
involved with Unix on current and new DEC hardware.

Currently, we are routing all Usenet traffic through duke.
If necessary or desirable, we could poll other systems as
well. For now, mail duke!decvax!person. Our network
names, etc. are below, feel free to contact us.

decvax!shannon - Bill Shannon (formerly of Case Western Reserve Univ.)
decvax!aps - Armando Stettner

This page last updated on: Jul 1 09:16