[TUHS] origin of null-terminated strings

[TUHS] origin of null-terminated strings

[Douglas McIlroy]

I think this cited quote from
https://www.joelonsoftware.com/2001/12/11/ is urban legend.

    Why do C strings [have a terminating NUl]? It’s because the PDP-7
microprocessor, on which UNIX and the C programming language were
invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
at the end.”

This assertion seems unlikely since neither C nor the library string
functions existed on the PDP-7. In fact the "terminating character" of
a string in the PDP-7 language B was the pair '*e'. A string was a
sequence of words, packed two characters per word. For odd-length
strings half of the final one-character word was effectively
NUL-padded as described below.

One might trace null termination to the original (1965) proposal for
ASCII,  https://dl.acm.org/doi/10.1145/363831.363839. There the only
role specifically suggested for NUL is to "serve to accomplish time
fill or media fill." With character-addressable hardware (not the
PDP-7), it is only a small step from using NUL as terminal padding to
the convention of null termination in all cases.

Ken would probably know for sure whether there's any  truth in the
attribution to ASCIZ.

Doug

[TUHS] Re: origin of null-terminated strings

[Ken Thompson]

[-- Attachment #1: Type: text/plain, Size: 1405 bytes --]

asciz -- this is the first time i heard of it.
doug -- yes.


On Thu, Dec 15, 2022 at 7:04 PM Douglas McIlroy <
douglas.mcilroy@dartmouth.edu> wrote:

> I think this cited quote from
> https://www.joelonsoftware.com/2001/12/11/ is urban legend.
>
>     Why do C strings [have a terminating NUl]? It’s because the PDP-7
> microprocessor, on which UNIX and the C programming language were
> invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
> at the end.”
>
> This assertion seems unlikely since neither C nor the library string
> functions existed on the PDP-7. In fact the "terminating character" of
> a string in the PDP-7 language B was the pair '*e'. A string was a
> sequence of words, packed two characters per word. For odd-length
> strings half of the final one-character word was effectively
> NUL-padded as described below.
>
> One might trace null termination to the original (1965) proposal for
> ASCII,  https://dl.acm.org/doi/10.1145/363831.363839. There the only
> role specifically suggested for NUL is to "serve to accomplish time
> fill or media fill." With character-addressable hardware (not the
> PDP-7), it is only a small step from using NUL as terminal padding to
> the convention of null termination in all cases.
>
> Ken would probably know for sure whether there's any  truth in the
> attribution to ASCIZ.
>
> Doug
>

[-- Attachment #2: Type: text/html, Size: 1982 bytes --]

[TUHS] Re: origin of null-terminated strings

[Dr Iain Maoileoin]

[-- Attachment #1: Type: text/plain, Size: 1969 bytes --]

ASCIZ
Lost in the mists of time in my mind.

I remember running into a .asciz directive n the 70s “somewhere”.
It was an assembler directive in one of the RT11 systems??? or perhaps the unix bootstrap and/or “.s” files - when I get some time I will go read some old code/manuals.

I

Yes, it put a null byte at the end of a string.

> On 16 Dec 2022, at 03:14, Ken Thompson <kenbob@gmail.com> wrote:
> 
> asciz -- this is the first time i heard of it.
> doug -- yes.
> 
> 
> On Thu, Dec 15, 2022 at 7:04 PM Douglas McIlroy <douglas.mcilroy@dartmouth.edu <mailto:douglas.mcilroy@dartmouth.edu>> wrote:
> I think this cited quote from
> https://www.joelonsoftware.com/2001/12/11/ <https://www.joelonsoftware.com/2001/12/11/> is urban legend.
> 
>     Why do C strings [have a terminating NUl]? It’s because the PDP-7
> microprocessor, on which UNIX and the C programming language were
> invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
> at the end.”
> 
> This assertion seems unlikely since neither C nor the library string
> functions existed on the PDP-7. In fact the "terminating character" of
> a string in the PDP-7 language B was the pair '*e'. A string was a
> sequence of words, packed two characters per word. For odd-length
> strings half of the final one-character word was effectively
> NUL-padded as described below.
> 
> One might trace null termination to the original (1965) proposal for
> ASCII,  https://dl.acm.org/doi/10.1145/363831.363839 <https://dl.acm.org/doi/10.1145/363831.363839>. There the only
> role specifically suggested for NUL is to "serve to accomplish time
> fill or media fill." With character-addressable hardware (not the
> PDP-7), it is only a small step from using NUL as terminal padding to
> the convention of null termination in all cases.
> 
> Ken would probably know for sure whether there's any  truth in the
> attribution to ASCIZ.
> 
> Doug


[-- Attachment #2: Type: text/html, Size: 3319 bytes --]

[TUHS] Re: origin of null-terminated strings

[Dan Halbert]

[-- Attachment #1: Type: text/plain, Size: 2436 bytes --]

ASCIZ was an assembler directive used for a number of different DEC 
computers, and also the name for null-terminated strings. I learned it 
for the PDP-10, but I'm sure it existed on other machines. It is in some 
PDP-10 documentation I am looking at right now. Anyone who used DEC and 
did assembly programming would have known about it. Various system calls 
took ASCIZ strings.

On 12/16/22 04:13, Dr Iain Maoileoin wrote:
> ASCIZ
> Lost in the mists of time in my mind.
>
> I remember running into a .asciz directive n the 70s “somewhere”.
> It was an assembler directive in one of the RT11 systems??? or perhaps 
> the unix bootstrap and/or “.s” files - when I get some time I will go 
> read some old code/manuals.
>
> I
>
> Yes, it put a null byte at the end of a string.
>
>> On 16 Dec 2022, at 03:14, Ken Thompson <kenbob@gmail.com> wrote:
>>
>> asciz -- this is the first time i heard of it.
>> doug -- yes.
>>
>>
>> On Thu, Dec 15, 2022 at 7:04 PM Douglas McIlroy 
>> <douglas.mcilroy@dartmouth.edu> wrote:
>>
>>     I think this cited quote from
>>     https://www.joelonsoftware.com/2001/12/11/ is urban legend.
>>
>>         Why do C strings [have a terminating NUl]? It’s because the PDP-7
>>     microprocessor, on which UNIX and the C programming language were
>>     invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z
>>     (zero)
>>     at the end.”
>>
>>     This assertion seems unlikely since neither C nor the library string
>>     functions existed on the PDP-7. In fact the "terminating
>>     character" of
>>     a string in the PDP-7 language B was the pair '*e'. A string was a
>>     sequence of words, packed two characters per word. For odd-length
>>     strings half of the final one-character word was effectively
>>     NUL-padded as described below.
>>
>>     One might trace null termination to the original (1965) proposal for
>>     ASCII, https://dl.acm.org/doi/10.1145/363831.363839. There the only
>>     role specifically suggested for NUL is to "serve to accomplish time
>>     fill or media fill." With character-addressable hardware (not the
>>     PDP-7), it is only a small step from using NUL as terminal padding to
>>     the convention of null termination in all cases.
>>
>>     Ken would probably know for sure whether there's any truth in the
>>     attribution to ASCIZ.
>>
>>     Doug
>>
>

[-- Attachment #2: Type: text/html, Size: 5515 bytes --]

[TUHS] Re: origin of null-terminated strings

[Dan Cross]

On Fri, Dec 16, 2022 at 8:42 AM Dan Halbert <halbert@halwitz.org> wrote:
> ASCIZ was an assembler directive used for a number of different DEC computers, and also the name for null-terminated strings. I learned it for the PDP-10, but I'm sure it existed on other machines. It is in some PDP-10 documentation I am looking at right now. Anyone who used DEC and did assembly programming would have known about it. Various system calls took ASCIZ strings.

This raises something I've always been curious about. To what extent were
the Unix folks at Bell Labs already familiar with DEC systems before the PDP-7?

It strikes me that much of the published work was centered around IBM and GE
systems (e.g., Ken's wonderful paper on regular expressions, and of course the
Multics work). Were there other Digital machines floating around? I know a
proposal was written to get a PDP-10 for operating systems research, but it
wasn't approved.

Relatedly, was any thought given to trying to get a 360 system?

On 12/16/22 04:13, Dr Iain Maoileoin wrote:
> ASCIZ
> Lost in the mists of time in my mind.

Origin, perhaps, but it exists in contemporary assemblers. Like most
sane people I try to avoid being in assembler for too long, when you're
first turning on a machine it is useful to be able to squirt a message
out of the UART if something goes dramatically wrong, and the directive
is handy for that.

It seems to have made its way into Research assembler via BSD; it's in
locore.s in 8th Edition, for instance, but doesn't appear before that.  The
"UNIX Assembler Manual" describes "String Statements" for the 7th
Edition assembler; strings are sequences of ASCII characters between
'<' and '>'.  But it doesn't say that they're NUL terminated, and they are
not: adding the terminator was manual via the familiar, `\0` escape
sequence.

        - Dan C.


> I remember running into a .asciz directive n the 70s “somewhere”.
> It was an assembler directive in one of the RT11 systems??? or perhaps the unix bootstrap and/or “.s” files - when I get some time I will go read some old code/manuals.
>
> I
>
> Yes, it put a null byte at the end of a string.
>
> On 16 Dec 2022, at 03:14, Ken Thompson <kenbob@gmail.com> wrote:
>
> asciz -- this is the first time i heard of it.
> doug -- yes.
>
>
> On Thu, Dec 15, 2022 at 7:04 PM Douglas McIlroy <douglas.mcilroy@dartmouth.edu> wrote:
>>
>> I think this cited quote from
>> https://www.joelonsoftware.com/2001/12/11/ is urban legend.
>>
>>     Why do C strings [have a terminating NUl]? It’s because the PDP-7
>> microprocessor, on which UNIX and the C programming language were
>> invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
>> at the end.”
>>
>> This assertion seems unlikely since neither C nor the library string
>> functions existed on the PDP-7. In fact the "terminating character" of
>> a string in the PDP-7 language B was the pair '*e'. A string was a
>> sequence of words, packed two characters per word. For odd-length
>> strings half of the final one-character word was effectively
>> NUL-padded as described below.
>>
>> One might trace null termination to the original (1965) proposal for
>> ASCII,  https://dl.acm.org/doi/10.1145/363831.363839. There the only
>> role specifically suggested for NUL is to "serve to accomplish time
>> fill or media fill." With character-addressable hardware (not the
>> PDP-7), it is only a small step from using NUL as terminal padding to
>> the convention of null termination in all cases.
>>
>> Ken would probably know for sure whether there's any  truth in the
>> attribution to ASCIZ.
>>
>> Doug
>
>
>

[TUHS] Re: origin of null-terminated strings

[Tom Lyon]

[-- Attachment #1: Type: text/plain, Size: 4327 bytes --]

Re: getting a 360 - IBM and AT&T really hated each other, so 360s were
avoided for strategic reasons. That said, they could not be practically
avoided; Holmdel had a large installation:
https://www.youtube.com/watch?v=HMYiktO0D64&ab_channel=AT%26TTechChannel

When Amdahl and UTS/UNIX came along, the Bell System was by far the biggest
customer.

On Fri, Dec 16, 2022 at 8:12 AM Dan Cross <crossd@gmail.com> wrote:

> On Fri, Dec 16, 2022 at 8:42 AM Dan Halbert <halbert@halwitz.org> wrote:
> > ASCIZ was an assembler directive used for a number of different DEC
> computers, and also the name for null-terminated strings. I learned it for
> the PDP-10, but I'm sure it existed on other machines. It is in some PDP-10
> documentation I am looking at right now. Anyone who used DEC and did
> assembly programming would have known about it. Various system calls took
> ASCIZ strings.
>
> This raises something I've always been curious about. To what extent were
> the Unix folks at Bell Labs already familiar with DEC systems before the
> PDP-7?
>
> It strikes me that much of the published work was centered around IBM and
> GE
> systems (e.g., Ken's wonderful paper on regular expressions, and of course
> the
> Multics work). Were there other Digital machines floating around? I know a
> proposal was written to get a PDP-10 for operating systems research, but it
> wasn't approved.
>
> Relatedly, was any thought given to trying to get a 360 system?
>
> On 12/16/22 04:13, Dr Iain Maoileoin wrote:
> > ASCIZ
> > Lost in the mists of time in my mind.
>
> Origin, perhaps, but it exists in contemporary assemblers. Like most
> sane people I try to avoid being in assembler for too long, when you're
> first turning on a machine it is useful to be able to squirt a message
> out of the UART if something goes dramatically wrong, and the directive
> is handy for that.
>
> It seems to have made its way into Research assembler via BSD; it's in
> locore.s in 8th Edition, for instance, but doesn't appear before that.  The
> "UNIX Assembler Manual" describes "String Statements" for the 7th
> Edition assembler; strings are sequences of ASCII characters between
> '<' and '>'.  But it doesn't say that they're NUL terminated, and they are
> not: adding the terminator was manual via the familiar, `\0` escape
> sequence.
>
>         - Dan C.
>
>
> > I remember running into a .asciz directive n the 70s “somewhere”.
> > It was an assembler directive in one of the RT11 systems??? or perhaps
> the unix bootstrap and/or “.s” files - when I get some time I will go read
> some old code/manuals.
> >
> > I
> >
> > Yes, it put a null byte at the end of a string.
> >
> > On 16 Dec 2022, at 03:14, Ken Thompson <kenbob@gmail.com> wrote:
> >
> > asciz -- this is the first time i heard of it.
> > doug -- yes.
> >
> >
> > On Thu, Dec 15, 2022 at 7:04 PM Douglas McIlroy <
> douglas.mcilroy@dartmouth.edu> wrote:
> >>
> >> I think this cited quote from
> >> https://www.joelonsoftware.com/2001/12/11/ is urban legend.
> >>
> >>     Why do C strings [have a terminating NUl]? It’s because the PDP-7
> >> microprocessor, on which UNIX and the C programming language were
> >> invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
> >> at the end.”
> >>
> >> This assertion seems unlikely since neither C nor the library string
> >> functions existed on the PDP-7. In fact the "terminating character" of
> >> a string in the PDP-7 language B was the pair '*e'. A string was a
> >> sequence of words, packed two characters per word. For odd-length
> >> strings half of the final one-character word was effectively
> >> NUL-padded as described below.
> >>
> >> One might trace null termination to the original (1965) proposal for
> >> ASCII,  https://dl.acm.org/doi/10.1145/363831.363839. There the only
> >> role specifically suggested for NUL is to "serve to accomplish time
> >> fill or media fill." With character-addressable hardware (not the
> >> PDP-7), it is only a small step from using NUL as terminal padding to
> >> the convention of null termination in all cases.
> >>
> >> Ken would probably know for sure whether there's any  truth in the
> >> attribution to ASCIZ.
> >>
> >> Doug
> >
> >
> >
>

[-- Attachment #2: Type: text/html, Size: 5568 bytes --]

[TUHS] Re: origin of null-terminated strings

[Jon Steinhart]

Dan Cross writes:
>
> This raises something I've always been curious about. To what extent were
> the Unix folks at Bell Labs already familiar with DEC systems before the PDP-7?

Well, I recall that there was a PDP-8 in the keypunch room on the 5th floor of
building 2.  I believe that it was hooked to a card reader and printer so that
one could get a listing of a deck of cards without having to use the computer
center.  But that's probably not what you're asking about.

Jon

[TUHS] Re: origin of null-terminated strings

[Dave Horsfall]

[-- Attachment #1: Type: text/plain, Size: 363 bytes --]

On Fri, 16 Dec 2022, Dr Iain Maoileoin wrote:

> I remember running into a .asciz directive n the 70s “somewhere”. It was 
> an assembler directive in one of the RT11 systems??? or perhaps the unix 
> bootstrap and/or “.s” files - when I get some time I will go read some 
> old code/manuals.

MACRO-11 on RSX-11D seems to ring a bell...

-- Dave

[TUHS] Re: origin of null-terminated strings

[Warner Losh]

[-- Attachment #1: Type: text/plain, Size: 639 bytes --]

On Fri, Dec 16, 2022, 1:12 PM Dave Horsfall <dave@horsfall.org> wrote:

> On Fri, 16 Dec 2022, Dr Iain Maoileoin wrote:
>
> > I remember running into a .asciz directive n the 70s “somewhere”. It was
> > an assembler directive in one of the RT11 systems??? or perhaps the unix
> > bootstrap and/or “.s” files - when I get some time I will go read some
> > old code/manuals.
>
> MACRO-11 on RSX-11D seems to ring a bell...
>

I first encountered it on RSTS/E 6C in the MACRO-11 it had... But the v6
macro assembler from DEC via Harvard that eventually wound up in 2BSD is
older and dates to 1977 or so.

Warner

>

[-- Attachment #2: Type: text/html, Size: 1179 bytes --]

[TUHS] Re: origin of null-terminated strings

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 2920 bytes --]

So I went to the oracle on much of DEC history ... -- this explains why Ken
never heard it.



---------- Forwarded message ---------
From: Timothe Litt
Date: Fri, Dec 16, 2022 at 3:40 PM
Subject: Re: Origin of ASCIZ / null terminated char arrays.
To: Clem Cole <clemc@ccc.com>

On 16-Dec-22 15:04, Clem Cole wrote:
Do either of you know when it showed up in DEC assemblers?  I  remember it
in Macro11 and Macro10, but I have to believe it was in the earlier
machines?  So far I have not found a reference to it in any of my PDP-8
stuff (which is small) and I never had the docs for 6, 7 or 9 -- I assume
Al K. has them on bitsavers - so I'm going to go poking around - but I
thought I'd ask you two if you knew.


Ken Thompson says he had never heard of it before, but he never used the
DEC assemblers -- (he wrote their own on the Honeywell originally I
believe). FWIW: B did not use null-terminated char arrays originally, but
by the time dmr morphed B into newB then C, they had become standard.  Like
many, I had always thought Dennis picked them from the DEC assembler, but
as Ken says - they never really used it.


I was trying to figure out when they (null terminate char arrays) started
to become more standard and specifically the pdeudo OP ASCIZ to create them.


Tx
Clem

It depends on if you require ASCII, or just character strings terminated by
a stop code...
The -11 has .asciz (as does VMS Macro,...); the -10 has ASCIZ.  SIXBIT 0 is
a space, so you needed to know the length, oftentimes in words, so strip
trailing 00s.
The basic 8 assembler (PAL) didn't even have ASCII data.
http://www.bitsavers.org/pdf/dec/pdp8/software/DEC-08-ASAC-D_PAL-III_Symbolic_Assembler_Programming_Manual.pdf

Macro-8 does; the TEXT pseudo-op uses 00 as a stop code.  (It also uses a
6-bit ASCII code).  " is a single character ASCII constant, but not used
for strings.
https://www.grc.com/pdp-8/docs/macro-8_programming_manual.pdf

The -15 has .ASCII and .SIXBIT, but no .ASCIZ.

http://bitsavers.informatik.uni-stuttgart.de/pdf/dec/pdp15/DEC-15-AMZA-D_MACRO15.pdf

Probably of most interest to the Unix history, the PDP-7 assembler's TEXT
pseudo-op 'in order to separate the string from other data following it, a
termination code determined by the character mode is inserted automatically
after the last character code of the string"/...

http://www.bitsavers.org/pdf/dec/pdp7/PDP-7_AsmMan.pdf
I don't remember and/or didn't use the earlier assemblers, but many of the
manuals are on bitsavers.
Both NUL and RUBOUT (a.k.a. DELETE) were used as fill characters to cover
the time teletypes take to execute <CR> and <LF>.  you couldn't represent
the NUL version with ASCIZ, and RUBOUT was picked for the ability to
overpunch paper tape typos.  Neither function, nor the use of NUL as an end
of string marker  is in the ASCII standard, IIRC.

ᐧ

> ᐧ

[-- Attachment #2: Type: text/html, Size: 5920 bytes --]

[TUHS] Re: origin of null-terminated strings

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 1002 bytes --]

More info WRT to historical DEC usage ...
---------- Forwarded message ---------
From: Bob Supnik
Date: Fri, Dec 16, 2022 at 4:39 PM
Subject: Re: Origin of ASCIZ / null terminated char arrays.
To: Clem Cole


It wasn't in the PDP8. The PDP8 mostly used sixbit, the ASCII subset
between 40 and 137. The character was simply masked by 077, so that 100
(@) became 0 and could be used as the delimiter. PAL8 (in OS8) does not
have a text generation pseudo-op.

The PDP7 had a TEXT pseudo-op that <did> fill an extra word with 0s if
the string was a multiple of 3 characters. It supported FIODEC, BAUDOT,
and ANALEX encodings, but not ASCII.

The PDP9 has both .SXBIT and .ASCII. The latter used two 18-bit words to
hold five 7bit ASCII characters. In both cases, words were zero-filled,
but an extra (word) of 0s was not added if the string was a multiple of
2/multiple of 5 characters.

The PDP11 had .ASCIZ, starting with Macro11 in 1972.

Tim can comment on the PDP10.

> ᐧ

[-- Attachment #2: Type: text/html, Size: 1647 bytes --]

[TUHS] Re: origin of null-terminated strings

[Phil Budne]

> From: Bob Supnik
> Tim can comment on the PDP10.

MACRO10 (the DEC PDP-10 assembler) had the ASCIZ directive,

I don't see it in the May 1964 MACRO6 (PDP-6 assembler) document at:

http://bitsavers.trailing-edge.com/pdf/dec/pdp6/F-64MAS_MACRO6_Assembly_Program_May64.pdf

Nor the February 1965 version:
http://bitsavers.trailing-edge.com/pdf/dec/pdp6/DEC-6-0-TP-MAC-LM-FP-ACT01_MACRO-6_Assembly_Language_Feb65.pdf

But it does appear in the May 1965 MACRO-6 manual:

http://bitsavers.trailing-edge.com/pdf/dec/pdp6/DEC-6-0-TP-MAC-LM-FP_ACT02_MACRO-6_Assembly_Language_May65.pdf

Which has the fullly trifuricated character packings:

ASCII/ASCIZ:	7 bit bytes, with the low order bit left over
		(set at the start of lines in files to indicate a Line
		Sequence Number metadata for line number based editors)
SIXBIT		"6-bit ASCII" -- ASCII characters 040 thru 0137
		stored as 00 thru 077 in six six bit bytes
RADIX50		6 characters from a 40 (050) character character set
		(plus four flag bits) used to store symbol tables
		https://en.wikipedia.org/wiki/DEC_RADIX_50#36-bit_systems

And ASCIZ is used in listings of the PDP-6 "T.S. Executive" version
1.4 dated 8-18-65:

http://bitsavers.trailing-edge.com/pdf/dec/pdp6/tsExec1.4/COMCON.pdf

COMCON is "COMmand CONtrol" -- the top level command interpreter built
into the monitor (the file name was retained into the later days of
TOPS-10), and messages output to the user use ASCIZ directives.

And to tie the thread back (closer) to the list subject, the "sub
title" headers in the above assembler listing file are "T. HASTINGS
8-2-65" (who I believe is Tom Hastings), which also appears in many
other files, including the job scheduler:

http://bitsavers.trailing-edge.com/pdf/dec/pdp6/tsExec1.4/CLKCSS.pdf

*AND* T. Hastings also appears as an author of the CTSS scheduler:

https://softwarehistory.csse.rose-hulman.edu/index.php/ctss-scheduler/
(in the "Full Code" section):

          :R******TIME SHARING SCHEDULING ALGORITHM***********
          :R    T. Hastings and R. Daley
          :R    Minor Modifications by G. Schroeder when NEW
          :R    I/O Package Installed....Summer, 1965

[TUHS] Re: origin of null-terminated strings

[Luther Johnson]

[-- Attachment #1: Type: text/plain, Size: 867 bytes --]

I used RT-11 versions 4 and 5, and I seem to remember the MACRO-11 there 
had .ASCIZ.

On 12/16/2022 02:02 PM, Warner Losh wrote:
>
>
> On Fri, Dec 16, 2022, 1:12 PM Dave Horsfall <dave@horsfall.org 
> <mailto:dave@horsfall.org>> wrote:
>
>     On Fri, 16 Dec 2022, Dr Iain Maoileoin wrote:
>
>     > I remember running into a .asciz directive n the 70s
>     “somewhere”. It was
>     > an assembler directive in one of the RT11 systems??? or perhaps
>     the unix
>     > bootstrap and/or “.s” files - when I get some time I will go
>     read some
>     > old code/manuals.
>
>     MACRO-11 on RSX-11D seems to ring a bell...
>
>
> I first encountered it on RSTS/E 6C in the MACRO-11 it had... But the 
> v6 macro assembler from DEC via Harvard that eventually wound up in 
> 2BSD is older and dates to 1977 or so.
>
> Warner


[-- Attachment #2: Type: text/html, Size: 1944 bytes --]

[TUHS] Re: origin of null-terminated strings

[Dan Halbert]

[-- Attachment #1: Type: text/plain, Size: 1370 bytes --]

On 12/16/22 16:02, Warner Losh wrote:
> On Fri, Dec 16, 2022, 1:12 PM Dave Horsfall <dave@horsfall.org> wrote:
>
>     On Fri, 16 Dec 2022, Dr Iain Maoileoin wrote:
>
>     > I remember running into a .asciz directive n the 70s
>     “somewhere”. It was
>     > an assembler directive in one of the RT11 systems??? or perhaps
>     the unix
>     > bootstrap and/or “.s” files - when I get some time I will go
>     read some
>     > old code/manuals.
>
>     MACRO-11 on RSX-11D seems to ring a bell...
>
>
> I first encountered it on RSTS/E 6C in the MACRO-11 it had... But the 
> v6 macro assembler from DEC via Harvard that eventually wound up in 
> 2BSD is older and dates to 1977 or so.
>
> Warner

The PDP-10 manual I spoke of is from 1971, and there were older 
editions. For the PDP-7, this manual from 1965, 
http://www.bitsavers.org/pdf/dec/pdp7/PDP-7_AsmMan.pdf, printed pages 
38-40, does not mention "ASCIZ" specifically, but talks about assembler 
directives "TELETYPE" and "ANALEX" that add a "termination code" of 00 
octal, for characters.

DEC also used SIXBIT, a truncated ASCII code that had printing 
characters but no control characters, so no newline, etc. In that 
scheme, 00 octal was SPACE. Table here: 
https://en.wikipedia.org/wiki/Six-bit_character_code#Examples_of_six-bit_ASCII_variants.

Dan H

[-- Attachment #2: Type: text/html, Size: 2677 bytes --]

[TUHS] Re: origin of null-terminated strings

[Steve Nickolas]

[-- Attachment #1: Type: text/plain, Size: 1910 bytes --]

On Thu, 15 Dec 2022, Douglas McIlroy wrote:

> I think this cited quote from
> https://www.joelonsoftware.com/2001/12/11/ is urban legend.
>
>    Why do C strings [have a terminating NUl]? It’s because the PDP-7
> microprocessor, on which UNIX and the C programming language were
> invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
> at the end.”
>
> This assertion seems unlikely since neither C nor the library string
> functions existed on the PDP-7. In fact the "terminating character" of
> a string in the PDP-7 language B was the pair '*e'. A string was a
> sequence of words, packed two characters per word. For odd-length
> strings half of the final one-character word was effectively
> NUL-padded as described below.
>
> One might trace null termination to the original (1965) proposal for
> ASCII,  https://dl.acm.org/doi/10.1145/363831.363839. There the only
> role specifically suggested for NUL is to "serve to accomplish time
> fill or media fill." With character-addressable hardware (not the
> PDP-7), it is only a small step from using NUL as terminal padding to
> the convention of null termination in all cases.
>
> Ken would probably know for sure whether there's any  truth in the
> attribution to ASCIZ.
>
> Doug
>

For what it's worth, when I code for the Apple //e (using 65C02 
assembler), I use C strings.  I can just do something like

prstr:   ldy       #$00
@1:      lda       msg, y
          beq       @2        ; string terminator
          ora       #$80      ; firmware wants high bit on
          jsr       $FDED     ; write char
          iny
          bne       @1
@2:      rts

msg:     .byte     "Hello, cruel world.", 13, 0

and using a NUL terminator just makes sense here because of how simple it 
is to check for (BEQ and BNE check the 6502's zero flag, which LDA 
automatically sets).

-uso.

[TUHS] Re: origin of null-terminated strings

[John P. Linderman]

[-- Attachment #1: Type: text/plain, Size: 4313 bytes --]

Suppose you have two strings of 8-bit bytes, and you'd like
to compare them lexicographically (left to right, byte by byte).
An oracle tells you the length of the strings, so maybe you have

3 ram
4 ramp

You can just do an strncmp on the two strings, using the minimum
of the two lengths (3 in the example). If they differ (they didn't),
you are done. If the strings are of the same length (they aren't),
they are equal. Otherwise, the shorter (a prefix of the longer)
compares low. Ho-hum.

Suppose each comparand is a sequence of such strings, and you
want to break ties on initial components of such sequences
using subsequent components (if any). But you have to combine
them as a single string and the oracle only tells you the total length.
You can't just concatenate them together, or

(3 ram, 4 part) => 7 rampart
(4 ramp, 3 art) => 7 rampart
(7 rampart) => 7 rampart

and they all look equal, but they're not supposed to be.
The problem is that some components are proper prefixes
of the corresponding component. We can sneak past the end
of one component and compare bytes from different components,
something that cannot be allowed. A collection of components
is said to have "the prefix property" if no component is
a proper prefix of any other component. If there is some
byte that cannot occur in some component, we can tack that
byte onto the component, and the components will then have
the prefix property. Newline terminated lines have the
prefix property, but we cannot just concatenate such
components and do an strncmp, because newlines compare low
to most bytes, but high to some legitimate bytes, like tabs,
so they don't break ties correctly.

Null terminators to the rescue! These induce the prefix property,
and compare low to everything else. Using blanks to stand in for
hard-to-parse null bytes, we get

(4 ram , 5 part ) => 9 ram part
(5 ramp , 4 art ) => 9 ramp art
(8 rampart )      => 8 rampart

The strncmp on the results establishes the desired order.
The null terminator on the final component is optional.
The oracular length determines how much of the component
is significant. But you have to be consistent. The final
component must always, or never, have the "optional" terminator.

Null-terminated strings (which cannot, themselves, contain a null)
are not the only components having the prefix property.
Fixed length strings cannot, by definition, include proper prefixes,
and they are free to contain any byte. UTF-8 code-points
have the prefix property (I suspect this was no accident),
so the null-terminated concatenation of non-null UTF-8
code-points have the prefix property and break ties appropriately
(assuming that the code-points themselves establish the
correct order for what is being compared).

I doubt that this explains the use of null terminated strings,
but for those of use who spend too much time thinking about sorting,
they sure work well.

On Thu, Dec 15, 2022 at 10:04 PM Douglas McIlroy <
douglas.mcilroy@dartmouth.edu> wrote:

> I think this cited quote from
> https://www.joelonsoftware.com/2001/12/11/ is urban legend.
>
>     Why do C strings [have a terminating NUl]? It’s because the PDP-7
> microprocessor, on which UNIX and the C programming language were
> invented, had an ASCIZ string type. ASCIZ meant “ASCII with a Z (zero)
> at the end.”
>
> This assertion seems unlikely since neither C nor the library string
> functions existed on the PDP-7. In fact the "terminating character" of
> a string in the PDP-7 language B was the pair '*e'. A string was a
> sequence of words, packed two characters per word. For odd-length
> strings half of the final one-character word was effectively
> NUL-padded as described below.
>
> One might trace null termination to the original (1965) proposal for
> ASCII,  https://dl.acm.org/doi/10.1145/363831.363839. There the only
> role specifically suggested for NUL is to "serve to accomplish time
> fill or media fill." With character-addressable hardware (not the
> PDP-7), it is only a small step from using NUL as terminal padding to
> the convention of null termination in all cases.
>
> Ken would probably know for sure whether there's any  truth in the
> attribution to ASCIZ.
>
> Doug
>

[-- Attachment #2: Type: text/html, Size: 5367 bytes --]

[TUHS] Terms for string, and similar character constructs (was: origin of null-terminated strings)

[Alejandro Colomar]

[Resend from my subscribed address, as the list is subscribers-only, it seems]

In C, most syscalls and libc functions use strings, that is, zero or more 
non-NUL characters followed by a NUL.

However, there are a few cases where other incompatible character constructs are 
used.  A few examples:

-  utmpx(5): Some of its fields use fixed-width char arrays which contain a 
sequence of non-NUL characters, and padding of NULs to fill the rest (although 
some systems only require a NUL to delimit the padding, which can then contain 
garbage).

-  Some programs use just a pointer and a length to determine sequences of 
characters.  No NULs involved.

-  abstract sockets:  On Linux, abstract Unix socket names are stored in a 
fixed-width array, and all bytes are meaningful (up to the specified size), even 
if they are NULs.  Only special that that the first byte is NUL.

Since those are only rare cases, those constructs don't seem to have a name; 
some programmers call them strings (quite confusingly).

Has there been any de-facto standard (or informal naming) to call those things, 
and differentiate them?


Thanks,

Alex

-- 
<http://www.alejandro-colomar.es/>

[TUHS] Re: Terms for string, and similar character constructs (was: origin of null-terminated strings)

[Dave Horsfall]

On Fri, 16 Dec 2022, Alejandro Colomar wrote:

> [Resend from my subscribed address, as the list is subscribers-only, it seems]

Of course; open lists are spam magnets...  Regretful, but true.

-- Dave

[TUHS] Re: origin of null-terminated strings

[Douglas McIlroy]

> To what extent were the Unix folks at Bell Labs already familiar with DEC systems before the PDP-7?

Some awareness, but no hands-on experience,

> was any thought given to trying to get a 360 system?

Very serious thought. However, virtual memory was a non-negotiable
desideratum, to which Gene Amdahl was implacably opposed because
demand paging would devastate hardware performance. Soon after GE got
the nod, IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
but by then the die had been cast. Michigan bought one and built a
nice time-sharing system that was running well before Multics.

Doug

[TUHS] Re: origin of null-terminated strings

[James Frew]

[-- Attachment #1: Type: text/plain, Size: 998 bytes --]

Yes, MTS (Michigan Terminal System)!

One of my first jobs as a geography grad student was helping Waldo 
Tobler <https://en.wikipedia.org/wiki/Waldo_R._Tobler> port a bunch of 
his MTS Fortran* programs to UCSB's 370/168 running OS/MVT. Definitely a 
step backwards, as Waldo never ceased to remind me.

He calmed down once he got a Tektronix storage tube display connected to 
our PDP-11/45 running v6; seemed like UNIX was almost as nice as MTS...

/Frew

*commented in German...

On 2022-12-16 14:26, Douglas McIlroy wrote:
>> was any thought given to trying to get a 360 system?
> Very serious thought. However, virtual memory was a non-negotiable
> desideratum, to which Gene Amdahl was implacably opposed because
> demand paging would devastate hardware performance. Soon after GE got
> the nod, IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
> but by then the die had been cast. Michigan bought one and built a
> nice time-sharing system that was running well before Multics.

[-- Attachment #2: Type: text/html, Size: 1651 bytes --]

[TUHS] Re: origin of null-terminated strings

[steve jenkin]

> On 17 Dec 2022, at 09:26, Douglas McIlroy <douglas.mcilroy@dartmouth.edu> wrote:
> 
>> 
>> was any thought given to trying to get a 360 system?
> 
> Very serious thought. However, virtual memory was a non-negotiable
> desideratum, to which Gene Amdahl was implacably opposed because
> demand paging would devastate hardware performance. Soon after GE got
> the nod, IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
> but by then the die had been cast. Michigan bought one and built a
> nice time-sharing system that was running well before Multics.
> 
> Doug

Doug,

Thanks for the insight.

I've seen “MTS” mentioned, but never properly understood its significance.
Never looked into it, either :(
Brief search results below, w/o Wikipedia etc.

A process step of building new ’things’, skipped on all computing / I.T. projects I worked on,
is a “Post Mortem” a.k.a “Post Implementation Review”.

If MIT / Bell Labs / GE ever did a project review on Multics, I’d love to know if it’s been published,
and what insights they came away with,
and any changes made to their development & project management processes.

The MIT lead, Corbató / Corby, had demonstrated a high-level of competence & ability.
He'd built CTSS in 1961 and won the ACM Turing Award in 1990. Never given to "second best”.

It wasn’t a lack of talent, need, desire (for a product/service to sell) or funding that made Multics take years & years.

With the capability & experience of the people involved, it'd be simplistic and superficial to attribute
the project delays to “Second System Effect”.

Is it more akin to what we’re seeing now in the differences in approach to building Space Launch Systems & Vehicles?

I don’t have words/ concepts for the different approaches, but they seem to parallel how Multics & Unix were developed.

	- NASA & Boeing et al have only just flown the Space Shuttle replacement, Artemis, the SLS pluse Orion capsule.
		In 2005, a program to replace the Shuttle (retired in 2011) was begun.
		This was replaced with the SLS / Artemis program in 2010 - reusing many of the Shuttle components, eg RS-25 engines.
		Next flight, Artemis 2, due in 2024.

	- Space-X is developing it’s second launch system, plus a Crew / Cargo vehicle (StarShip).
		Falcon 9 has become the cheapest per kg/LEO, most reliable and most flown rocket in history.
		They’ve already beaten 1 launch/week this year, lofting 150+ tons per quarter into LEO.
		Which is more than two-times all other programs, public or private, put together.

steve j

========

 David L. Mills, photo gallery & some comments

	Michigan Terminal System (MTS)
		<https://www.eecis.udel.edu/~mills/gallery/gallery8.html>

		During much of the 1960s I was a staff member at the U Michigan Computing Center.
		I worked with a bunch of other guys on various hardward and software projects, 
			one of which is described on this page.

========

Organization and features of the Michigan terminal system
	Michael T. Alexander
	16 November 1971
	<https://dl.acm.org/doi/abs/10.1145/1478873.1478951>

	ABSTRACT

		This paper will explore some aspects of the Michigan Terminal System (MTS) developed at the University of Michigan. 
		MTS is the operating system used on the IBM 360/67 at the University of Michigan Computing Center,
			 as well as at several other installations.

		It supports a large variety of uses ranging from very small student-type jobs to large jobs requiring several million bytes of storage and hours of processor time. 
		Currently at the University of Michigan there are about 13,000 users running as many as 86,000 jobs per month.

========

Time-sharing in the IBM system/360: model 67
	Charles T. Gibson
	IBM
	<https://dl.acm.org/doi/abs/10.1145/1464182.1464190>

		Fine detail of 360/67, differences to /50.
		TSS mentioned

========

--
Steve Jenkin, IT Systems and Design 
0412 786 915 (+61 412 786 915)
PO Box 38, Kippax ACT 2615, AUSTRALIA

mailto:sjenkin@canb.auug.org.au http://members.tip.net.au/~sjenkin

[TUHS] Re: origin of null-terminated strings

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 2883 bytes --]

Given the number of ex-MTS (Bill Joy and Ted Kowalski, to name two) and TSS
hackers that were also later to be UNIX hackers after their original
introduction to system programming as undergrads.  I will keep this reply
in TUHS, although it could be argued that it belongs in COFF.

Note good sources for even more of the background of the history politics
at both IBM & GE can be found in Haigh and Ceruzzi's book: "A New History
of Modern Computing
<https://www.amazon.com/New-History-Modern-Computing/dp/0262542900>" -
which I have previously mentioned as it is a beautiful read.

On Fri, Dec 16, 2022 at 5:27 PM Douglas McIlroy <
douglas.mcilroy@dartmouth.edu> wrote:

> IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
> but by then the die had been cast. Michigan bought one and built a
> nice time-sharing system that was running well before Multics.
>
All true, but a few details are glossed over, and thus, this could be
misinterpreted - so I'm going to add those as one of the people.

TSS and the /67 was IBM's answer to Multics, as Doug mentions.  Note that the
/67 could run as a model /65, which as I understand it, most of the ones
IBM sold did.

At the time, IBM offered the /67 to Universities at a substantial discount
(I believe even less than the /65).  Thus, several schools bought them with
Michigan, CMU, Cornell, and Princeton that I am aware of; but I suspect
there were others.

TSS was late, and the first releases could have been more stable.   Cornell
and Princeton chose to run their systems as /65 using the original IBM OS.
CMU and Michigan both received copies of TSS with their systems.   Michigan
would do a substantial rewrite, which was different enough that became the
new system MTS.   CMU did a great deal of bug fixing, which went back to
IBM, and they chose to run TSS.  [I believe that CMU runs OS/360 by data
and TSS at night until they felt they could trust it to not crash].
Nominally, TSS and MTS should share programs, and with some work, both
could import source programs from OS/360 [My first paid programming job was
helping to rewrite York/APL from OS/360 to run on TSS].  So the compilers
and many tools for all three were common.

MTS and TSS used the same file system structure, or it was close enough
that tools were shared.  I don't know if OS/360 could read TSS disk packs -
I would have suspected, although the common media of the day was 1/2" mag
tape.

This leads to a UNIX legacy that ...  Ted's fsck(8) - which purists know as
a different name in the first version -  was modeled after the disk
scavenger program from TSS and MTS.   icheck/ncheck et al. seem pretty
primitive if you had used to see the other as a system programmer first.
 Also, a big reason why all the errors were originally in uppercase was the
IBM program had done it.  In many ways, neither Ted nor I knew any better
at the time.

Clem

[-- Attachment #2: Type: text/html, Size: 5131 bytes --]

[TUHS] Re: origin of null-terminated strings

[Tom Lyon]

[-- Attachment #1: Type: text/plain, Size: 3454 bytes --]

Clem doesn't mention CP-67/CMS, which IBM kept trying to kill in favor of
CMS.
From Melinda Varian's amazing history of VM, I gleaned these factoids:
CP-67 - 8 sites by May '68
Feb of 68 - IBM decommits from TSS
Apr 69 - IBM rescinds decommit of TSS
CP-67 - 44 sites by 1970, ~10 internal to IBM
May 71 - TSS finally decommitted

So TSS was a rocky road, while CP&VM were simple and just worked.



On Sat, Dec 17, 2022 at 9:13 AM Clem Cole <clemc@ccc.com> wrote:

> Given the number of ex-MTS (Bill Joy and Ted Kowalski, to name two) and
> TSS hackers that were also later to be UNIX hackers after their original
> introduction to system programming as undergrads.  I will keep this reply
> in TUHS, although it could be argued that it belongs in COFF.
>
> Note good sources for even more of the background of the history politics
> at both IBM & GE can be found in Haigh and Ceruzzi's book: "A New History
> of Modern Computing
> <https://www.amazon.com/New-History-Modern-Computing/dp/0262542900>" -
> which I have previously mentioned as it is a beautiful read.
>
> On Fri, Dec 16, 2022 at 5:27 PM Douglas McIlroy <
> douglas.mcilroy@dartmouth.edu> wrote:
>
>> IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
>> but by then the die had been cast. Michigan bought one and built a
>> nice time-sharing system that was running well before Multics.
>>
> All true, but a few details are glossed over, and thus, this could be
> misinterpreted - so I'm going to add those as one of the people.
>
> TSS and the /67 was IBM's answer to Multics, as Doug mentions.  Note that the
> /67 could run as a model /65, which as I understand it, most of the ones
> IBM sold did.
>
> At the time, IBM offered the /67 to Universities at a substantial discount
> (I believe even less than the /65).  Thus, several schools bought them with
> Michigan, CMU, Cornell, and Princeton that I am aware of; but I suspect
> there were others.
>
> TSS was late, and the first releases could have been more stable.
>  Cornell and Princeton chose to run their systems as /65 using the original
> IBM OS.  CMU and Michigan both received copies of TSS with their systems.
>  Michigan would do a substantial rewrite, which was different enough that
> became the new system MTS.   CMU did a great deal of bug fixing, which went
> back to IBM, and they chose to run TSS.  [I believe that CMU runs OS/360 by
> data and TSS at night until they felt they could trust it to not crash].
> Nominally, TSS and MTS should share programs, and with some work, both
> could import source programs from OS/360 [My first paid programming job was
> helping to rewrite York/APL from OS/360 to run on TSS].  So the compilers
> and many tools for all three were common.
>
> MTS and TSS used the same file system structure, or it was close enough
> that tools were shared.  I don't know if OS/360 could read TSS disk packs -
> I would have suspected, although the common media of the day was 1/2" mag
> tape.
>
> This leads to a UNIX legacy that ...  Ted's fsck(8) - which purists know
> as a different name in the first version -  was modeled after the disk
> scavenger program from TSS and MTS.   icheck/ncheck et al. seem pretty
> primitive if you had used to see the other as a system programmer first.
>  Also, a big reason why all the errors were originally in uppercase was the
> IBM program had done it.  In many ways, neither Ted nor I knew any better
> at the time.
>
> Clem
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 6030 bytes --]

[TUHS] Re: origin of null-terminated strings

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 5161 bytes --]

Tom Lyon -- TSS was around and supported into the 80's.  That said, I've
seen that May '71, but it might be a typo -- '81 sounds much more plausible
as it real death.   IIRC Tom Haight has better dates in his book.

FWIW: I was at CMU in the mid 70s [programming TSS including installing
fixes from the IBM support team].  Plus, my old boss, Dean Hiller, left CMU
in the late 70s to work for IBM as a TSS system person [he retired from IBM
years later and had moved to the AIX team at one point].   And I also have
a copy of one of the TSS documents that has a printing date of 1980.

It's also possible IBM stopped *selling new sites* in the early 70s,  but
TSS was definitely a supported product throughout the 1980s.  IBM had some
large and important customers running TSS, in particular, NASA and I
believe a couple of automotive ones -- maybe GM and Rolls Royce but I don't
know.   IIRC: One of the original mechanical CAD programs had been
developed on it and users needed either MTS or TSS to run it properly.

I also remember that in 77-78, when CMU started to move off the /67 to the
DEC-20s, IBM had counter-proposed an S370/168 with VM on it - which CMU had
rejected.  But Amdahl had proposed CMU could keep running TSS on their
then-newest system which was at least the V7 (maybe the V8 as I have
forgotten when the latter was released).

Around that same time, Michigan had stayed with MTS but had switched to
Amdhal as the vendor.
ᐧ

On Sat, Dec 17, 2022 at 1:15 PM Tom Lyon <pugs78@gmail.com> wrote:

> Clem doesn't mention CP-67/CMS, which IBM kept trying to kill in favor of
> CMS.
> From Melinda Varian's amazing history of VM, I gleaned these factoids:
> CP-67 - 8 sites by May '68
> Feb of 68 - IBM decommits from TSS
> Apr 69 - IBM rescinds decommit of TSS
> CP-67 - 44 sites by 1970, ~10 internal to IBM
> May 71 - TSS finally decommitted
>
> So TSS was a rocky road, while CP&VM were simple and just worked.
>
>
>
> On Sat, Dec 17, 2022 at 9:13 AM Clem Cole <clemc@ccc.com> wrote:
>
>> Given the number of ex-MTS (Bill Joy and Ted Kowalski, to name two) and
>> TSS hackers that were also later to be UNIX hackers after their original
>> introduction to system programming as undergrads.  I will keep this reply
>> in TUHS, although it could be argued that it belongs in COFF.
>>
>> Note good sources for even more of the background of the history politics
>> at both IBM & GE can be found in Haigh and Ceruzzi's book: "A New
>> History of Modern Computing
>> <https://www.amazon.com/New-History-Modern-Computing/dp/0262542900>" -
>> which I have previously mentioned as it is a beautiful read.
>>
>> On Fri, Dec 16, 2022 at 5:27 PM Douglas McIlroy <
>> douglas.mcilroy@dartmouth.edu> wrote:
>>
>>> IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
>>> but by then the die had been cast. Michigan bought one and built a
>>> nice time-sharing system that was running well before Multics.
>>>
>> All true, but a few details are glossed over, and thus, this could be
>> misinterpreted - so I'm going to add those as one of the people.
>>
>> TSS and the /67 was IBM's answer to Multics, as Doug mentions.  Note that
>> the /67 could run as a model /65, which as I understand it, most of the
>> ones IBM sold did.
>>
>> At the time, IBM offered the /67 to Universities at a
>> substantial discount (I believe even less than the /65).  Thus, several
>> schools bought them with Michigan, CMU, Cornell, and Princeton that I am
>> aware of; but I suspect there were others.
>>
>> TSS was late, and the first releases could have been more stable.
>>  Cornell and Princeton chose to run their systems as /65 using the original
>> IBM OS.  CMU and Michigan both received copies of TSS with their systems.
>>  Michigan would do a substantial rewrite, which was different enough that
>> became the new system MTS.   CMU did a great deal of bug fixing, which went
>> back to IBM, and they chose to run TSS.  [I believe that CMU runs OS/360 by
>> data and TSS at night until they felt they could trust it to not crash].
>> Nominally, TSS and MTS should share programs, and with some work, both
>> could import source programs from OS/360 [My first paid programming job was
>> helping to rewrite York/APL from OS/360 to run on TSS].  So the compilers
>> and many tools for all three were common.
>>
>> MTS and TSS used the same file system structure, or it was close enough
>> that tools were shared.  I don't know if OS/360 could read TSS disk packs -
>> I would have suspected, although the common media of the day was 1/2" mag
>> tape.
>>
>> This leads to a UNIX legacy that ...  Ted's fsck(8) - which purists know
>> as a different name in the first version -  was modeled after the disk
>> scavenger program from TSS and MTS.   icheck/ncheck et al. seem pretty
>> primitive if you had used to see the other as a system programmer first.
>>  Also, a big reason why all the errors were originally in uppercase was the
>> IBM program had done it.  In many ways, neither Ted nor I knew any better
>> at the time.
>>
>> Clem
>>
>>
>>
>>

[-- Attachment #2: Type: text/html, Size: 8923 bytes --]

[TUHS] Re: origin of null-terminated strings

[Clem Cole]

[-- Attachment #1: Type: text/plain, Size: 5465 bytes --]

typo... sigh...

TSS was definitely a supported product throughout *the 1970s and into* the
1980s
ᐧ

On Sat, Dec 17, 2022 at 1:43 PM Clem Cole <clemc@ccc.com> wrote:

> Tom Lyon -- TSS was around and supported into the 80's.  That said, I've
> seen that May '71, but it might be a typo -- '81 sounds much more plausible
> as it real death.   IIRC Tom Haight has better dates in his book.
>
> FWIW: I was at CMU in the mid 70s [programming TSS including installing
> fixes from the IBM support team].  Plus, my old boss, Dean Hiller, left CMU
> in the late 70s to work for IBM as a TSS system person [he retired from IBM
> years later and had moved to the AIX team at one point].   And I also have
> a copy of one of the TSS documents that has a printing date of 1980.
>
> It's also possible IBM stopped *selling new sites* in the early 70s,  but
> TSS was definitely a supported product throughout the 1980s.  IBM had some
> large and important customers running TSS, in particular, NASA and I
> believe a couple of automotive ones -- maybe GM and Rolls Royce but I don't
> know.   IIRC: One of the original mechanical CAD programs had been
> developed on it and users needed either MTS or TSS to run it properly.
>
> I also remember that in 77-78, when CMU started to move off the /67 to the
> DEC-20s, IBM had counter-proposed an S370/168 with VM on it - which CMU had
> rejected.  But Amdahl had proposed CMU could keep running TSS on their
> then-newest system which was at least the V7 (maybe the V8 as I have
> forgotten when the latter was released).
>
> Around that same time, Michigan had stayed with MTS but had switched to
> Amdhal as the vendor.
> ᐧ
>
> On Sat, Dec 17, 2022 at 1:15 PM Tom Lyon <pugs78@gmail.com> wrote:
>
>> Clem doesn't mention CP-67/CMS, which IBM kept trying to kill in favor of
>> CMS.
>> From Melinda Varian's amazing history of VM, I gleaned these factoids:
>> CP-67 - 8 sites by May '68
>> Feb of 68 - IBM decommits from TSS
>> Apr 69 - IBM rescinds decommit of TSS
>> CP-67 - 44 sites by 1970, ~10 internal to IBM
>> May 71 - TSS finally decommitted
>>
>> So TSS was a rocky road, while CP&VM were simple and just worked.
>>
>>
>>
>> On Sat, Dec 17, 2022 at 9:13 AM Clem Cole <clemc@ccc.com> wrote:
>>
>>> Given the number of ex-MTS (Bill Joy and Ted Kowalski, to name two) and
>>> TSS hackers that were also later to be UNIX hackers after their original
>>> introduction to system programming as undergrads.  I will keep this reply
>>> in TUHS, although it could be argued that it belongs in COFF.
>>>
>>> Note good sources for even more of the background of the history
>>> politics at both IBM & GE can be found in Haigh and Ceruzzi's book: "A
>>> New History of Modern Computing
>>> <https://www.amazon.com/New-History-Modern-Computing/dp/0262542900>" -
>>> which I have previously mentioned as it is a beautiful read.
>>>
>>> On Fri, Dec 16, 2022 at 5:27 PM Douglas McIlroy <
>>> douglas.mcilroy@dartmouth.edu> wrote:
>>>
>>>> IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
>>>> but by then the die had been cast. Michigan bought one and built a
>>>> nice time-sharing system that was running well before Multics.
>>>>
>>> All true, but a few details are glossed over, and thus, this could be
>>> misinterpreted - so I'm going to add those as one of the people.
>>>
>>> TSS and the /67 was IBM's answer to Multics, as Doug mentions.  Note
>>> that the /67 could run as a model /65, which as I understand it, most
>>> of the ones IBM sold did.
>>>
>>> At the time, IBM offered the /67 to Universities at a
>>> substantial discount (I believe even less than the /65).  Thus, several
>>> schools bought them with Michigan, CMU, Cornell, and Princeton that I am
>>> aware of; but I suspect there were others.
>>>
>>> TSS was late, and the first releases could have been more stable.
>>>  Cornell and Princeton chose to run their systems as /65 using the original
>>> IBM OS.  CMU and Michigan both received copies of TSS with their systems.
>>>  Michigan would do a substantial rewrite, which was different enough that
>>> became the new system MTS.   CMU did a great deal of bug fixing, which went
>>> back to IBM, and they chose to run TSS.  [I believe that CMU runs OS/360 by
>>> data and TSS at night until they felt they could trust it to not crash].
>>> Nominally, TSS and MTS should share programs, and with some work, both
>>> could import source programs from OS/360 [My first paid programming job was
>>> helping to rewrite York/APL from OS/360 to run on TSS].  So the compilers
>>> and many tools for all three were common.
>>>
>>> MTS and TSS used the same file system structure, or it was close enough
>>> that tools were shared.  I don't know if OS/360 could read TSS disk packs -
>>> I would have suspected, although the common media of the day was 1/2" mag
>>> tape.
>>>
>>> This leads to a UNIX legacy that ...  Ted's fsck(8) - which purists know
>>> as a different name in the first version -  was modeled after the disk
>>> scavenger program from TSS and MTS.   icheck/ncheck et al. seem pretty
>>> primitive if you had used to see the other as a system programmer first.
>>>  Also, a big reason why all the errors were originally in uppercase was the
>>> IBM program had done it.  In many ways, neither Ted nor I knew any better
>>> at the time.
>>>
>>> Clem
>>>
>>>
>>>
>>>

[-- Attachment #2: Type: text/html, Size: 9979 bytes --]

[TUHS] Re: origin of null-terminated strings

[Tom Perrine]

[-- Attachment #1: Type: text/plain, Size: 9322 bytes --]

I can offer at least one perspective on Multics, although from a
limited viewpoint.

I was an Explorer Scout at post 414, sponsored by Honeywell, in Phoenix,
from about 1972 onwards. This was a "high tech" special interest Post, and
our interest was computers - no surprise.

Weekly meetings included time using Teletypes (and later nicer terminals)
to connect to a GCOS system in Phoenix. Programming in Basic, FORTRAN,
COBOL and other arcane languages ensued. As did "an incident". Pretty
minor, but yeah, some high school kids "popped" (for some value of
"popped") the local GCOS system and got access to some files they shouldn't
have. I'm not sure if this was dumpster diving, or a real bug, or just
exploiting some employee carelessly setting file permissions to "read and
write for the entire system". I strongly suspect the latter.

As a result, all our GCOS accounts were cancelled and we were moved to
System M, one of the primary development systems for Multics. As was new
and not all common at the time, the entire source code for the entire
system was available online - and not just because this was a dev system,
this was the norm for all Multics owners. In this way, Multics followed the
UNIX model, or maybe they came from a common theme.

The Post took to PL/1 in a big way and delivered some interesting changes
to (mostly) the games on the system - vbg (Video BackGammon), chess, yet
another space wars clone, and a huge expanded ADVENTURE. There was also
some interesting new programming, including finding a way to exploit IPC
channels, and another way to crash the front end processor, that led to
some emergency bugfixes being rolled out (I later learned) overnight to the
Pentagon and "the Fort" customers.

Speaking of which, there was a bit of a kerfluffle when someone who shall
remain nameless, innocently asked "I identified all the Multics systems in
the site documents, except System N. Where is System N?" in a public forum
on System M. I was warned Never To Speak of System "N" Ever Again Or Else.

Over the next 10 years, a lot of my technical life was around Multic. As a
Boy Scout, and eventually as a student intern at Honeywell. Throughout
college all my engineering programming was done on Multics, mostly in pl1,
instead of fortran. As an intern I worked on projects across the GCOS
(GCOS-8), Multics and even a little CP-6 spaces. I got to see and hear how
different parts of the company saw the GCOS vs Multics conflict, admittedly
filtered through the eyes of a still naive intern. Here are some
recollections and talking points.


   1. Multics hardware is too expensive!
      1. GCOS 8 HW, which would have run Multics, GCOS classic and the
      someday GCOS 8 was even more expensive)
      2. Multics HW was expensive because they built and burned-in an
      entire GCOS system, and then partially took it apart, re-wired the CPU
      boards and other parts, and then put it back together!A Multics CPU had
      some boards in common with a GCOS CPU, some were modified
boards, and some
      were completely separate, if I recall correctly. Wirewrap CPU boards in
      addition to the wirewrapped backplane, which was also (I think) a
      built-and-then-rewired GCOS backplane.
   2. Our customers want more batch and less timesharing! - Multics doesn't
   have a real batch mode!
      1. which could have easily been addressed with a "dusty deck
      translator" from GCOS JCL to Multics "absentee" scripts.
   3. Timesharing is a fad and too expensive - look at how few people  our
   GCOS customers put on TSS!
      1. yeah, because GCOS TSS was an afterthought - a big pasted-on batch
      job that remained primitive compared to Multics and other competitors -
      looking at you TOPS-20, which was also up and coming
   4.  Our profit margins on GCOS SW and HW are better - so we should sell
   more GCOS!
   5. Spending money on Multics will take away resources from our bread and
   butter GCOS, which is where we make money!
   6. No one cares about PL/1 - all our customers want COBOL and FORTRAN
      1. which Multics had, and a better COBOL compiler than GCOS - I
      learned COBOL68 and COBOL74 on both - and along the way found a bug that
      would crash THE ENTIRE GCOS OPERATING SYSTEM (not just the compiler!) if
      you mis-spelled "ENVIRONMENT DIVISION" in a COBOL74 program.
   7. GCOS (formerly GECOS, thankyouverymuch) is the real legacy of GE, not
   that research project from MIT!!! - Multics is a toy we were paid to make,
   that will never make a dime!
   8. There was some rivalry between Phoenix and Billerica(?) that was
   mostly kept away from "the kids", so we didn't see that in detail but it
   was there.
   9. This rivalry seemed to be (in retrospect) two orgs fighting over
   ever-scarcer resources as the mini/super-mini revolution came on strong. A
   Multics CPU was pretty much a synchronous 1 MIP machine (per CPU). Of
   course you could have up to 8(?) CPUs and the I/O bandwidth was light years
   ahead of the super-minis, but DEC said that the VaX-11/780 was a "1 MIP
   machine" and it was cheaper than a mainframe, so there!!

Anything else would probably be off topic for TUHS, but with the
inter-twined DNA of Multics and UNIX, perhaps this is actually on-topic?

I would point out that I don't think Multics was a victim of UNIX adoption,
as that (from my dim recollection) took place years after Multics dev had
been mostly capped.

After Multics, my first UNIX system was PWB on a PDP-11. It was quite the
change, but led to the rest of my career - BSD, KSOS, SysV, IRIX, Ultrix,
SunOS, Solaris, UNICOS, others and eventually Linux.

You never forget your first :-)

Regards,
--tep




On Sat, Dec 17, 2022 at 10:17 AM Tom Lyon <pugs78@gmail.com> wrote:

> Clem doesn't mention CP-67/CMS, which IBM kept trying to kill in favor of
> CMS.
> From Melinda Varian's amazing history of VM, I gleaned these factoids:
> CP-67 - 8 sites by May '68
> Feb of 68 - IBM decommits from TSS
> Apr 69 - IBM rescinds decommit of TSS
> CP-67 - 44 sites by 1970, ~10 internal to IBM
> May 71 - TSS finally decommitted
>
> So TSS was a rocky road, while CP&VM were simple and just worked.
>
>
>
> On Sat, Dec 17, 2022 at 9:13 AM Clem Cole <clemc@ccc.com> wrote:
>
>> Given the number of ex-MTS (Bill Joy and Ted Kowalski, to name two) and
>> TSS hackers that were also later to be UNIX hackers after their original
>> introduction to system programming as undergrads.  I will keep this reply
>> in TUHS, although it could be argued that it belongs in COFF.
>>
>> Note good sources for even more of the background of the history politics
>> at both IBM & GE can be found in Haigh and Ceruzzi's book: "A New
>> History of Modern Computing
>> <https://www.amazon.com/New-History-Modern-Computing/dp/0262542900>" -
>> which I have previously mentioned as it is a beautiful read.
>>
>> On Fri, Dec 16, 2022 at 5:27 PM Douglas McIlroy <
>> douglas.mcilroy@dartmouth.edu> wrote:
>>
>>> IBM revealed Gerrit Blaauw's skunk-works project, the 360/67,
>>> but by then the die had been cast. Michigan bought one and built a
>>> nice time-sharing system that was running well before Multics.
>>>
>> All true, but a few details are glossed over, and thus, this could be
>> misinterpreted - so I'm going to add those as one of the people.
>>
>> TSS and the /67 was IBM's answer to Multics, as Doug mentions.  Note that
>> the /67 could run as a model /65, which as I understand it, most of the
>> ones IBM sold did.
>>
>> At the time, IBM offered the /67 to Universities at a
>> substantial discount (I believe even less than the /65).  Thus, several
>> schools bought them with Michigan, CMU, Cornell, and Princeton that I am
>> aware of; but I suspect there were others.
>>
>> TSS was late, and the first releases could have been more stable.
>>  Cornell and Princeton chose to run their systems as /65 using the original
>> IBM OS.  CMU and Michigan both received copies of TSS with their systems.
>>  Michigan would do a substantial rewrite, which was different enough that
>> became the new system MTS.   CMU did a great deal of bug fixing, which went
>> back to IBM, and they chose to run TSS.  [I believe that CMU runs OS/360 by
>> data and TSS at night until they felt they could trust it to not crash].
>> Nominally, TSS and MTS should share programs, and with some work, both
>> could import source programs from OS/360 [My first paid programming job was
>> helping to rewrite York/APL from OS/360 to run on TSS].  So the compilers
>> and many tools for all three were common.
>>
>> MTS and TSS used the same file system structure, or it was close enough
>> that tools were shared.  I don't know if OS/360 could read TSS disk packs -
>> I would have suspected, although the common media of the day was 1/2" mag
>> tape.
>>
>> This leads to a UNIX legacy that ...  Ted's fsck(8) - which purists know
>> as a different name in the first version -  was modeled after the disk
>> scavenger program from TSS and MTS.   icheck/ncheck et al. seem pretty
>> primitive if you had used to see the other as a system programmer first.
>>  Also, a big reason why all the errors were originally in uppercase was the
>> IBM program had done it.  In many ways, neither Ted nor I knew any better
>> at the time.
>>
>> Clem
>>
>>
>>
>>

[-- Attachment #2: Type: text/html, Size: 12557 bytes --]

[TUHS] Re: origin of null-terminated strings

[Adam Thornton]

> On Dec 17, 2022, at 11:15 AM, Tom Lyon <pugs78@gmail.com> wrote:
> 
> Clem doesn't mention CP-67/CMS, which IBM kept trying to kill in favor of CMS.
> From Melinda Varian's amazing history of VM, I gleaned these factoids:
> CP-67 - 8 sites by May '68
> Feb of 68 - IBM decommits from TSS
> Apr 69 - IBM rescinds decommit of TSS
> CP-67 - 44 sites by 1970, ~10 internal to IBM
> May 71 - TSS finally decommitted
> 
> So TSS was a rocky road, while CP&VM were simple and just worked.
> 

Ask the Varians!  Lee was a TSS heavy-hitter, and Melinda was...well, Melinda; "What Mother Never Told You" is still the best document about system maintenance ever written.

At least as of last year they were both still alive and well.

Adam

[TUHS] Re: origin of null-terminated strings

[Noel Chiappa]

    > From: Bob Supnik

    > The PDP11 had .ASCIZ, starting with Macro11 in 1972.

I was just about to report on my results, after a tiny bit of digging, which
included this. The important datum is that PAL-11 (in DEC-11-GGPB-D, "paper
tape software", April 1970, revised March 1971), which _preceded_ Macro-11,
_does not_ include .ASCIZ (although it has .ASCII). My oldest Macro-11 book
(DEC-11-OMACA-B-D, "BATCH-11/DOS-11 Assembler (MACRO-11)", April 1972, revised
March 1973) does have .ASCIZ. So in the DEC PDP-11 universe, it dates from
sometime between 1970 and 1972.

I'm not sure if Bell had any of the DEC paper tape software: "In early 1970 we
proposed acquisition of a PDP-11, which had just been introduced by
Digital. ... an order for a PDP-11 was placed in May. The processor arrived at
the end of the summer, but the PDP-11 was so new a product that no disk was
available until December. In the meantime, a rudimentary, core-only version of
Unix was written using a cross-assembler on the PDP-7." So the .ASCIZ in
Macro-11 wasn't until a couple of years later.

	 Noel