See the second column on the first page:

https://dl.acm.org/doi/pdf/10.1145/365230.365233

Louis

Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
> See the second column on the first page:
>
> https://dl.acm.org/doi/pdf/10.1145/365230.365233

Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
that stream procedures bypassed all the hardware protection mechanisms. I
think Burroughs fixed this on later machines. ALGOL was later sunk by the
development of ALGOL-68, which (AIUI) was more different from ALGOL-60 than
C++ is from C.

ALGOL and PL/I suffered from the same syndrome where FORTRAN programmers
wanted FORTRAN and COBOL programmers wanted COBOL, and that didn’t leave
much room for anything else at the time.

--
Pete

On Thursday, September 1, 2022 at 4:13:14 PM UTC-6, Peter Flass wrote:
> ALGOL was later sunk by the
> development of ALGOL-68, which (AIUI) was more different from ALGOL-60 than
> C++ is from C.

I have to agree that Algol-68 sank Algol. I had thought that the reason for that,
aside from silly things like case ... esac, was that implementing Algol-68 was
somewhat beyond the state of the art at the time.

That doesn't mean that there were _no_ implementations, just that many who
might have implemented an update to Algol did not feel themselves qualified
to produce an adequate implementation.

Of course, the world could have just ignored Algol-68, and stuck with Algol-60,
perhaps with slight tweaks and improvements. What closed off _that_ line of
development was...

Pascal.

John Savard

On Friday, September 2, 2022 at 7:18:46 AM UTC+10, Louis Krupp wrote:
> See the second column on the first page:
>
> https://dl.acm.org/doi/pdf/10.1145/365230.365233
..
The "over-the-horizon language" due mid-1966 was probably PL/I,
that considerably extended both ALGOL-60 and FORTRAN.
..
ALGOL was a particularly good medium for expressing readable numerical algorithms,
and it was a pity that some authors submitted programs in a less-well endowed language.

On 9/1/2022 4:13 PM, Peter Flass wrote:
> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>> See the second column on the first page:
>>
>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
> that stream procedures bypassed all the hardware protection mechanisms. I
> think Burroughs fixed this on later machines. ALGOL was later sunk by the
> development of ALGOL-68, which (AIUI) was more different from ALGOL-60 than
> C++ is from C.

Stream procedures could manipulate words that had the flag bit set, and
to the best of my recollection, that's the only hardware protection the
B5500 had. The Large System series -- the B6500/6700 on up -- had a more
complete capability architecture with three tag bits, and compilers that
generated user programs wouldn't emit operators that would touch words
with odd tags.

As a teenage nerd, I thought stream procedures were the coolest thing ever.

If Burroughs Large Systems -- or their Unisys successors -- had
attracted more users, especially at universities, there might have been
a reason to design a practical and portable version of ALGOL. Burroughs
Extended ALGOL and Large Systems' architecture were said to have been
designed together; I attended a few Burroughs users' conferences until
about 1982, and portability of source programs to other systems was a
concept I never heard discussed.

I would guess that most users of Burroughs Extended ALGOL had no idea
that the language had a history outside Burroughs and that things called
ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.

> ALGOL and PL/I suffered from the same syndrome where FORTRAN programmers
> wanted FORTRAN and COBOL programmers wanted COBOL, and that didn’t leave
> much room for anything else at the time.
>

There was a PL/I compiler for Burroughs Large Systems, but it was big
and complex and slow and not entirely bug-free and relatively few people
used it.

Louis

Louis Krupp <lkrupp@invalid.pssw.com.invalid> schrieb:
> See the second column on the first page:
>
> https://dl.acm.org/doi/pdf/10.1145/365230.365233

Lack of standardized I/O was a huge design flaw, each compiler
had to roll its own, so there was no portability across systems.

Kernighan wrote this in "UNIX: A History and a Memoir" that there
are three possibilities of doing I/O in a langue: Integrated into
the language itself (Fortran, Pascal), implemented in a library (C)
or not specifyling this at all. The last he called (from memory)
the least desirable option.

And FORTRAN was not only groundbreaking in optimization and writing
formulas in a "natural" way. Its I/O was also radical and new,
the FORMAT statement was a big innovation.

On Friday, September 2, 2022 at 4:11:52 PM UTC+10, Louis Krupp wrote:
> On 9/1/2022 4:13 PM, Peter Flass wrote:
> > Louis Krupp <lkr...@invalid.pssw.com.invalid> wrote:
> >> See the second column on the first page:
> >>
> >> https://dl.acm.org/doi/pdf/10.1145/365230.365233
> > Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
> > that stream procedures bypassed all the hardware protection mechanisms. I
> > think Burroughs fixed this on later machines. ALGOL was later sunk by the
> > development of ALGOL-68, which (AIUI) was more different from ALGOL-60 than
> > C++ is from C.
> Stream procedures could manipulate words that had the flag bit set, and
> to the best of my recollection, that's the only hardware protection the
> B5500 had. The Large System series -- the B6500/6700 on up -- had a more
> complete capability architecture with three tag bits, and compilers that
> generated user programs wouldn't emit operators that would touch words
> with odd tags.
>
> As a teenage nerd, I thought stream procedures were the coolest thing ever.
>
> If Burroughs Large Systems -- or their Unisys successors -- had
> attracted more users, especially at universities, there might have been
> a reason to design a practical and portable version of ALGOL. Burroughs
> Extended ALGOL and Large Systems' architecture were said to have been
> designed together; I attended a few Burroughs users' conferences until
> about 1982, and portability of source programs to other systems was a
> concept I never heard discussed.
>
> I would guess that most users of Burroughs Extended ALGOL had no idea
> that the language had a history outside Burroughs and that things called
> ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
> > ALGOL and PL/I suffered from the same syndrome where FORTRAN programmers
> > wanted FORTRAN and COBOL programmers wanted COBOL, and that didn’t leave
> > much room for anything else at the time.
>
> There was a PL/I compiler for Burroughs Large Systems, but it was big
> and complex and slow and not entirely bug-free and relatively few people
> used it.
..
My PL/I string-processing program, which I sent to another site, worked first time.

On Friday, September 2, 2022 at 4:50:06 PM UTC+10, Thomas Koenig wrote:
> Louis Krupp <lkr...@invalid.pssw.com.invalid> schrieb:
..
> Lack of standardized I/O was a huge design flaw, each compiler
> had to roll its own, so there was no portability across systems.

Indeed. Lack of portability was a difficulty. And it wasn't just the
I/O that wasn't portable.
..
> Kernighan wrote this in "UNIX: A History and a Memoir" that there
> are three possibilities of doing I/O in a langue: Integrated into
> the language itself (Fortran, Pascal), implemented in a library (C)
> or not specifyling this at all. The last he called (from memory)
> the least desirable option.
>
> And FORTRAN was not only groundbreaking in optimization and writing
> formulas in a "natural" way. Its I/O was also radical and new,
> the FORMAT statement was a big innovation.
..
Various ALGOLs implemented something that was just as good,
namely, to specify the number of digits in the printing of the number.

On 9/1/2022 11:11 PM, Louis Krupp wrote:
> On 9/1/2022 4:13 PM, Peter Flass wrote:
>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>>> See the second column on the first page:
>>>
>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
>> that stream procedures bypassed all the hardware protection mechanisms. I
>> think Burroughs fixed this on later machines. ALGOL was later sunk by the
>> development of ALGOL-68, which (AIUI) was more different from ALGOL-60
>> than
>> C++ is from C.
>
> Stream procedures could manipulate words that had the flag bit set, and
> to the best of my recollection, that's the only hardware protection the
> B5500 had. The Large System series -- the B6500/6700 on up -- had a more
> complete capability architecture with three tag bits, and compilers that
> generated user programs wouldn't emit operators that would touch words
> with odd tags.
>
> As a teenage nerd, I thought stream procedures were the coolest thing ever.
>
> If Burroughs Large Systems -- or their Unisys successors -- had
> attracted more users, especially at universities, there might have been
> a reason to design a practical and portable version of ALGOL.

Given that the original Algol didn't specify any I/O, so each
implementation did their own, it was hard to be portable. In fact one
"definition" of Pascal, was Algol plus I/O.

> Burroughs
> Extended ALGOL and Large Systems' architecture were said to have been
> designed together; I attended a few Burroughs users' conferences until
> about 1982, and portability of source programs to other systems was a
> concept I never heard discussed.
>
> I would guess that most users of Burroughs Extended ALGOL had no idea
> that the language had a history outside Burroughs and that things called
> ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.

There were Algol compilers for many different architectures. For
example, the Univac 1100 series actually had two different ones. Lack
of compilers wasn't a major factor in Algol's lack of popularity.

>> ALGOL and PL/I suffered from the same syndrome where FORTRAN programmers
>> wanted FORTRAN and COBOL programmers wanted COBOL, and that didn’t leave
>> much room for anything else at the time.
>>
>
> There was a PL/I compiler for Burroughs Large Systems, but it was big
> and complex and slow and not entirely bug-free and relatively few people
> used it.

Same for other architectures. Pl/1 was/is a large and complex language.
Early compilers didn't do too well. I believe it never gained much
popularity outside of IBM, and even there its success was "modest".

--
- Stephen Fuld
(e-mail address disguised to prevent spam)

On Thu, 1 Sep 2022 17:58:51 -0700 (PDT)
Quadibloc <jsavard@ecn.ab.ca> wrote:

> On Thursday, September 1, 2022 at 4:13:14 PM UTC-6, Peter Flass wrote:
> > ALGOL was later sunk by the
> > development of ALGOL-68, which (AIUI) was more different from ALGOL-60
> > than C++ is from C.
>
> I have to agree that Algol-68 sank Algol. I had thought that the reason
> for that, aside from silly things like case ... esac, was that
> implementing Algol-68 was somewhat beyond the state of the art at the
> time.

That and the whole business of stropping (by case or underline
or ...) was strange.

> That doesn't mean that there were _no_ implementations, just that many who
> might have implemented an update to Algol did not feel themselves
> qualified to produce an adequate implementation.

Algol 68C had quite a following at Cambridge circa 1980 - I suspect
that following dwindled when C was finally allowed at Cambridge. Some of
the software for the Torch CP/M machine was written in Algol 68C and
compiled to Z80 code on the University's 370. There was even a public
reading (aloud) of "The Revised Report on the Programming Language Algol
68" by a group of students (not including me - I thought the language to
be hideously overcomplex, but then I liked BCPL!.

Fun piece of folklore, the Algol 68C project was initially led by
Stephen Bourne of Bourne shell fame.

At around the same time Algol-W was the language used for teaching a
lot of things, including the details of how subroutines and local variables
behaved written in baroque terms based around text substitutions in the
source code.

--
Steve O'Hara-Smith
Odds and Ends at http://www.sohara.org/

Ahem A Rivet's Shot <steveo@eircom.net> schrieb:

> Fun piece of folklore, the Algol 68C project was initially led by
> Stephen Bourne of Bourne shell fame.

Looking at the syntax of the Bourne shell (and the style it was
written originally, with #defines making C into something much
resembling Algol) I certainly can believe that.

On 9/2/2022 9:00 AM, Stephen Fuld wrote:
> On 9/1/2022 11:11 PM, Louis Krupp wrote:
>> On 9/1/2022 4:13 PM, Peter Flass wrote:
>>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>>>> See the second column on the first page:
>>>>
>>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58,
>>> except
>>> that stream procedures bypassed all the hardware protection
>>> mechanisms. I
>>> think Burroughs fixed this on later machines. ALGOL was later sunk
>>> by the
>>> development of ALGOL-68, which (AIUI) was more different from
>>> ALGOL-60 than
>>> C++ is from C.
>>
>> Stream procedures could manipulate words that had the flag bit set,
>> and to the best of my recollection, that's the only hardware
>> protection the B5500 had. The Large System series -- the B6500/6700
>> on up -- had a more complete capability architecture with three tag
>> bits, and compilers that generated user programs wouldn't emit
>> operators that would touch words with odd tags.
>>
>> As a teenage nerd, I thought stream procedures were the coolest thing
>> ever.
>>
>> If Burroughs Large Systems -- or their Unisys successors -- had
>> attracted more users, especially at universities, there might have
>> been a reason to design a practical and portable version of ALGOL.
>
> Given that the original Algol didn't specify any I/O, so each
> implementation did their own, it was hard to be portable.  In fact one
> "definition" of Pascal, was Algol plus I/O.

Burroughs Extended ALGOL borrowed formatting from FORTRAN, changing
FORMAT statements to FORMAT declarations. It seemed natural and obvious
at the time; it would be surprising -- and unfortunate -- if no other
ALGOL implementations did something similar.

>
>
>> Burroughs Extended ALGOL and Large Systems' architecture were said to
>> have been designed together; I attended a few Burroughs users'
>> conferences until about 1982, and portability of source programs to
>> other systems was a concept I never heard discussed.
>>
>> I would guess that most users of Burroughs Extended ALGOL had no idea
>> that the language had a history outside Burroughs and that things
>> called ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
>
> There were Algol compilers for many different architectures.  For
> example, the Univac 1100 series actually had two different ones. Lack
> of compilers wasn't a major factor in Algol's lack of popularity.

Burroughs users tended to be an insular bunch, and I was no exception. I
recall one users conference attendee saying "If my firm changes systems,
I'll change firms." I never imagined that I'd ever be working on
anything else.

<snip>

Louis

On 9/2/2022 11:05 AM, Louis Krupp wrote:
> On 9/2/2022 9:00 AM, Stephen Fuld wrote:
>> On 9/1/2022 11:11 PM, Louis Krupp wrote:
>>> On 9/1/2022 4:13 PM, Peter Flass wrote:
>>>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>>>>> See the second column on the first page:
>>>>>
>>>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>>> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58,
>>>> except
>>>> that stream procedures bypassed all the hardware protection
>>>> mechanisms. I
>>>> think Burroughs fixed this on later machines. ALGOL was later sunk
>>>> by the
>>>> development of ALGOL-68, which (AIUI) was more different from
>>>> ALGOL-60 than
>>>> C++ is from C.
>>>
>>> Stream procedures could manipulate words that had the flag bit set,
>>> and to the best of my recollection, that's the only hardware
>>> protection the B5500 had. The Large System series -- the B6500/6700
>>> on up -- had a more complete capability architecture with three tag
>>> bits, and compilers that generated user programs wouldn't emit
>>> operators that would touch words with odd tags.
>>>
>>> As a teenage nerd, I thought stream procedures were the coolest thing
>>> ever.
>>>
>>> If Burroughs Large Systems -- or their Unisys successors -- had
>>> attracted more users, especially at universities, there might have
>>> been a reason to design a practical and portable version of ALGOL.
>>
>> Given that the original Algol didn't specify any I/O, so each
>> implementation did their own, it was hard to be portable.  In fact one
>> "definition" of Pascal, was Algol plus I/O.
>
> Burroughs Extended ALGOL borrowed formatting from FORTRAN, changing
> FORMAT statements to FORMAT declarations. It seemed natural and obvious
> at the time; it would be surprising -- and unfortunate -- if no other
> ALGOL implementations did something similar.

I don't know about any other Algol implementations, but Pascal certainly
made a different choice.

>>
>>> Burroughs Extended ALGOL and Large Systems' architecture were said to
>>> have been designed together; I attended a few Burroughs users'
>>> conferences until about 1982, and portability of source programs to
>>> other systems was a concept I never heard discussed.
>>>
>>> I would guess that most users of Burroughs Extended ALGOL had no idea
>>> that the language had a history outside Burroughs and that things
>>> called ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
>>
>> There were Algol compilers for many different architectures.  For
>> example, the Univac 1100 series actually had two different ones. Lack
>> of compilers wasn't a major factor in Algol's lack of popularity.
>
> Burroughs users tended to be an insular bunch, and I was no exception. I
> recall one users conference attendee saying "If my firm changes systems,
> I'll change firms." I never imagined that I'd ever be working on
> anything else.

I think that wasn't uncommon with programmers of many different systems.
The mainframes, and even mini-computer systems from different
manufacturers were so different from each other that once you spent a
lot of time and energy learning one, you were reluctant to "waste" all
the learning.

--
- Stephen Fuld
(e-mail address disguised to prevent spam)

On 02/09/2022 19:16, Stephen Fuld wrote:
> On 9/2/2022 11:05 AM, Louis Krupp wrote:
>> Burroughs Extended ALGOL borrowed formatting from FORTRAN, changing
>> FORMAT statements to FORMAT declarations. It seemed natural and
>> obvious at the time; it would be surprising -- and unfortunate --
>> if no other ALGOL implementations did something similar.
> I don't know about any other Algol implementations, but Pascal
> certainly made a different choice.

Algol went from essentially nothing in Algol 60 to a full-
blown I/O model including heavy-duty formats [inc declarations and
variables] in Algol 68.

I suspect that both Pascal and Algol 68 would have looked at
least somewhat different if their designers could have peeked into the
future and seen how C did it -- and if the designers of C could have
peeked into the past to see how Algol did it after that peek. [Sorry
about the recursion!]

I'll leave Pascal to its enthusiasts [if any], but one of the
problems [and perhaps the most serious] with Algol 68 was the fact
that formats were introduced as a whole new raft of syntax [somewhere
around a third of the entire language]. C showed that formatted I/O
could be done with no new syntax, just a handful of procedure calls
built around a small number of primitives and using strings for the
formats; further, the C model of files as streams of bytes is much
simpler than the Algol model of books with pages and lines [no matter
how well that matched the hardware of the period]. If Algol had
learned that lesson from C, C might then have learned from Algol
how to do unformatted I/O more easily. Win-win. But we can't go
back in time that way.

--
Andy Walker, Nottingham.
Andy's music pages: www.cuboid.me.uk/andy/Music
Composer of the day: www.cuboid.me.uk/andy/Music/Composers/Morel

Thomas Koenig <tkoenig@netcologne.de> wrote:
> Louis Krupp <lkrupp@invalid.pssw.com.invalid> schrieb:
>> See the second column on the first page:
>>
>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>
> Lack of standardized I/O was a huge design flaw, each compiler
> had to roll its own, so there was no portability across systems.

On the other hand, in reality simple I/O is pretty common across systems:
read a line/write a line, etc. It’s only the more exotic operations that
differ. If I were writing an ALGOL compiler I’d look at what the other guys
did and imitate it, as compatibly as possible.

>
> Kernighan wrote this in "UNIX: A History and a Memoir" that there
> are three possibilities of doing I/O in a langue: Integrated into
> the language itself (Fortran, Pascal), implemented in a library (C)
> or not specifyling this at all. The last he called (from memory)
> the least desirable option.
>
> And FORTRAN was not only groundbreaking in optimization and writing
> formulas in a "natural" way. Its I/O was also radical and new,
> the FORMAT statement was a big innovation.
>

--
Pete

Robin Vowels <robin.vowels@gmail.com> wrote:
> On Friday, September 2, 2022 at 4:11:52 PM UTC+10, Louis Krupp wrote:
>> On 9/1/2022 4:13 PM, Peter Flass wrote:
>>> Louis Krupp <lkr...@invalid.pssw.com.invalid> wrote:
>>>> See the second column on the first page:
>>>>
>>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
>>> that stream procedures bypassed all the hardware protection mechanisms. I
>>> think Burroughs fixed this on later machines. ALGOL was later sunk by the
>>> development of ALGOL-68, which (AIUI) was more different from ALGOL-60 than
>>> C++ is from C.
>> Stream procedures could manipulate words that had the flag bit set, and
>> to the best of my recollection, that's the only hardware protection the
>> B5500 had. The Large System series -- the B6500/6700 on up -- had a more
>> complete capability architecture with three tag bits, and compilers that
>> generated user programs wouldn't emit operators that would touch words
>> with odd tags.
>>
>> As a teenage nerd, I thought stream procedures were the coolest thing ever.
>>
>> If Burroughs Large Systems -- or their Unisys successors -- had
>> attracted more users, especially at universities, there might have been
>> a reason to design a practical and portable version of ALGOL. Burroughs
>> Extended ALGOL and Large Systems' architecture were said to have been
>> designed together; I attended a few Burroughs users' conferences until
>> about 1982, and portability of source programs to other systems was a
>> concept I never heard discussed.
>>
>> I would guess that most users of Burroughs Extended ALGOL had no idea
>> that the language had a history outside Burroughs and that things called
>> ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
>>> ALGOL and PL/I suffered from the same syndrome where FORTRAN programmers
>>> wanted FORTRAN and COBOL programmers wanted COBOL, and that didn’t leave
>>> much room for anything else at the time.
>>
>> There was a PL/I compiler for Burroughs Large Systems, but it was big
>> and complex and slow and not entirely bug-free and relatively few people
>> used it.
> .
> My PL/I string-processing program, which I sent to another site, worked first time.
>

PL/I is pretty compatible. I regularly compile programs written in PL/I(F)
and 2.3 with Iron Spring PL/I with minimal changes. OS/2 and Windows PL/I
took some of the I/O off in a different direction, so I have a couple of
things I wrote for OS/2 that are going to need some more work.

--
Pete

Stephen Fuld <sfuld@alumni.cmu.edu.invalid> wrote:
> On 9/1/2022 11:11 PM, Louis Krupp wrote:
>> On 9/1/2022 4:13 PM, Peter Flass wrote:
>>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>>>> See the second column on the first page:
>>>>
>>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
>>> that stream procedures bypassed all the hardware protection mechanisms. I
>>> think Burroughs fixed this on later machines. ALGOL was later sunk by the
>>> development of ALGOL-68, which (AIUI) was more different from ALGOL-60
>>> than
>>> C++ is from C.
>>
>> Stream procedures could manipulate words that had the flag bit set, and
>> to the best of my recollection, that's the only hardware protection the
>> B5500 had. The Large System series -- the B6500/6700 on up -- had a more
>> complete capability architecture with three tag bits, and compilers that
>> generated user programs wouldn't emit operators that would touch words
>> with odd tags.
>>
>> As a teenage nerd, I thought stream procedures were the coolest thing ever.
>>
>> If Burroughs Large Systems -- or their Unisys successors -- had
>> attracted more users, especially at universities, there might have been
>> a reason to design a practical and portable version of ALGOL.
>
> Given that the original Algol didn't specify any I/O, so each
> implementation did their own, it was hard to be portable. In fact one
> "definition" of Pascal, was Algol plus I/O.
>
>
>> Burroughs
>> Extended ALGOL and Large Systems' architecture were said to have been
>> designed together; I attended a few Burroughs users' conferences until
>> about 1982, and portability of source programs to other systems was a
>> concept I never heard discussed.
>>
>> I would guess that most users of Burroughs Extended ALGOL had no idea
>> that the language had a history outside Burroughs and that things called
>> ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
>
> There were Algol compilers for many different architectures. For
> example, the Univac 1100 series actually had two different ones. Lack
> of compilers wasn't a major factor in Algol's lack of popularity.

Univac went completely off the deep end. They had at least two 1100 COBOL
compilers, too.

--
Pete

Peter Flass wrote:
> Stephen Fuld <sfuld@alumni.cmu.edu.invalid> wrote:
>> On 9/1/2022 11:11 PM, Louis Krupp wrote:
>>> On 9/1/2022 4:13 PM, Peter Flass wrote:
>>>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>>>>> See the second column on the first page:
>>>>>
>>>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>>> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
>>>> that stream procedures bypassed all the hardware protection mechanisms. I
>>>> think Burroughs fixed this on later machines. ALGOL was later sunk by the
>>>> development of ALGOL-68, which (AIUI) was more different from ALGOL-60
>>>> than
>>>> C++ is from C.
>>>
>>> Stream procedures could manipulate words that had the flag bit set, and
>>> to the best of my recollection, that's the only hardware protection the
>>> B5500 had. The Large System series -- the B6500/6700 on up -- had a more
>>> complete capability architecture with three tag bits, and compilers that
>>> generated user programs wouldn't emit operators that would touch words
>>> with odd tags.
>>>
>>> As a teenage nerd, I thought stream procedures were the coolest thing ever.
>>>
>>> If Burroughs Large Systems -- or their Unisys successors -- had
>>> attracted more users, especially at universities, there might have been
>>> a reason to design a practical and portable version of ALGOL.
>>
>> Given that the original Algol didn't specify any I/O, so each
>> implementation did their own, it was hard to be portable. In fact one
>> "definition" of Pascal, was Algol plus I/O.
>>
>>
>>> Burroughs
>>> Extended ALGOL and Large Systems' architecture were said to have been
>>> designed together; I attended a few Burroughs users' conferences until
>>> about 1982, and portability of source programs to other systems was a
>>> concept I never heard discussed.
>>>
>>> I would guess that most users of Burroughs Extended ALGOL had no idea
>>> that the language had a history outside Burroughs and that things called
>>> ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
>>
>> There were Algol compilers for many different architectures. For
>> example, the Univac 1100 series actually had two different ones. Lack
>> of compilers wasn't a major factor in Algol's lack of popularity.
>
> Univac went completely off the deep end. They had at least two 1100 COBOL
> compilers, too.
>

That's an outsiders view, if you knew the reasons it made perfect sense.

The 1100/2200 side has two addressing modes, the original one dating
back to the post-Noah's Ark cleanup (Basic Mode), and one which I first
noticed in the 1980s (Extended Mode, aka UCS). UCS allows you to access
far more memory than BM does, BM has no problems with instructions but
the amount of data you can have visible is really limited by today's
standards.

Univac also started off with a 6-bit character set called Fieldata.

- COB (or FCOB not sure) - BM, Fieldata
- ACOB - BM, Ascii but can handle Fieldata
- UCOB - EM/UCS, Ascii with rudimentary Fdata capabilities
- OO-COB - EM/UCS with an IDE, I never used it.
The various compilers also implement different COBOL standards, I
believe OO-COB no longer understands the ALTER command for instance.

Interaction between BM and EM/UCS is "difficult", switching between the
two requires the use of MASM (assembler). MASM is also the only program
which can generate BM code and UCS code, or a mixture of the two.

Converting ACOB programs to UCOB can be trivial or it can be fraught -
depending on the use of certain features by the programmer, there is
also a compatability mode you can use.

On Saturday, September 3, 2022 at 1:00:36 AM UTC+10, Stephen Fuld wrote:
> On 9/1/2022 11:11 PM, Louis Krupp wrote:
> > On 9/1/2022 4:13 PM, Peter Flass wrote:
> >> Louis Krupp <lkr...@invalid.pssw.com.invalid> wrote:
> >>> See the second column on the first page:
> >>>
> >>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
> >> Yes, and? B5500 ALGOL was a very good implementation of ALGOL-58, except
> >> that stream procedures bypassed all the hardware protection mechanisms.. I
> >> think Burroughs fixed this on later machines. ALGOL was later sunk by the
> >> development of ALGOL-68, which (AIUI) was more different from ALGOL-60
> >> than
> >> C++ is from C.
> >
> > Stream procedures could manipulate words that had the flag bit set, and
> > to the best of my recollection, that's the only hardware protection the
> > B5500 had. The Large System series -- the B6500/6700 on up -- had a more
> > complete capability architecture with three tag bits, and compilers that
> > generated user programs wouldn't emit operators that would touch words
> > with odd tags.
> >
> > As a teenage nerd, I thought stream procedures were the coolest thing ever.
> >
> > If Burroughs Large Systems -- or their Unisys successors -- had
> > attracted more users, especially at universities, there might have been
> > a reason to design a practical and portable version of ALGOL.
> Given that the original Algol didn't specify any I/O, so each
> implementation did their own, it was hard to be portable. In fact one
> "definition" of Pascal, was Algol plus I/O.
> > Burroughs
> > Extended ALGOL and Large Systems' architecture were said to have been
> > designed together; I attended a few Burroughs users' conferences until
> > about 1982, and portability of source programs to other systems was a
> > concept I never heard discussed.
> >
> > I would guess that most users of Burroughs Extended ALGOL had no idea
> > that the language had a history outside Burroughs and that things called
> > ALGOL-58 and ALGOL-60 existed. I certainly didn't have a clue.
> There were Algol compilers for many different architectures. For
> example, the Univac 1100 series actually had two different ones. Lack
> of compilers wasn't a major factor in Algol's lack of popularity.
> >> ALGOL and PL/I suffered from the same syndrome where FORTRAN programmers
> >> wanted FORTRAN and COBOL programmers wanted COBOL, and that didn’t leave
> >> much room for anything else at the time.
> >>
> >
> > There was a PL/I compiler for Burroughs Large Systems, but it was big
> > and complex and slow and not entirely bug-free and relatively few people
> > used it.
> Same for other architectures. Pl/1 was/is a large and complex language.
..
On the contrary, the rules were simpler than for FORTRAN.
1. functions were generic. No need to bother with single and double precision
versions of the arithmetic and trig functions.
2. DO-loops had no restrictions on initial, final, and increment values;
no subset of expressions for these entities;
and loops could be executed zero or more times.
3. No restrictions on the complexity of subscript values;
4. Dynamic arrays -- no mucking about with "adjustable dimensions".
5. Clean interpretation of format specifications.
6. expressions allowed for field width of format items, etc
7. recursion allowed.
8. Real character strings, not fakes ones known as Hollerith constants.
9. Not necessary to count the characters in character constants.
10. arguments automatically checked for compatibility with corresponding
parameters on procedure calls.
11. arguments that are constants cannot be corrupted by assignment to
the corresponding parameter.
12. Single _AND_ double-precision complex.
There's more.
..
> Early compilers didn't do too well. I believe it never gained much
> popularity outside of IBM, and even there its success was "modest".
..
PL/C was a pretty good implementation, and fast .
Other non-IBM PL/I compilers included those for
CDC, Univac, Burroughs, and on micros DR PL/I, Q1/Lite PL/I.

Robin Vowels <robin.vowels@gmail.com> wrote:

> PL/C was a pretty good implementation, and fast .
> Other non-IBM PL/I compilers included those for
> CDC, Univac, Burroughs, and on micros DR PL/I, Q1/Lite PL/I.
>

Bitsavers has almost nothing on the Q1/Lite. if you, or anyone else, has
anything on this system it would be interesting to upload it.

--
Pete

On 9/2/2022 11:05 AM, Louis Krupp wrote:
> On 9/2/2022 9:00 AM, Stephen Fuld wrote:
>> On 9/1/2022 11:11 PM, Louis Krupp wrote:
>>> On 9/1/2022 4:13 PM, Peter Flass wrote:
>>>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> wrote:
>>>>> See the second column on the first page:
>>>>>
>>>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>>> Yes, and? B5500 ALGOL was a very good implementation of
>>>> ALGOL-58, except that stream procedures bypassed all the
>>>> hardware protection mechanisms. I think Burroughs fixed this on
>>>> later machines. ALGOL was later sunk by the development of
>>>> ALGOL-68, which (AIUI) was more different from ALGOL-60 than
>>>> C++ is from C.

No, B5000 and B5500 ALGOL were implementations of ALGOL-60, not
ALGOL-58, but the hardware and compiler were designed and implemented
before the Revised Report was published in January 1963.

Burroughs 220 BALGOL was closely based on ALGOL-58 (although it wasn't
called that originally -- it was the International Algorithmic Language,
or IAL). ALGOL-58 was never a real language design, just a progress
report on the work that eventually produced ALGOL-60.
>>>
>>> Stream procedures could manipulate words that had the flag bit
>>> set, and to the best of my recollection, that's the only hardware
>>> protection the B5500 had. The Large System series -- the
>>> B6500/6700 on up -- had a more complete capability architecture
>>> with three tag bits, and compilers that generated user programs
>>> wouldn't emit operators that would touch words with odd tags.
>>>
>>> As a teenage nerd, I thought stream procedures were the coolest
>>> thing ever.

Yes, Stream Procedures were very cool, and very useful, and very, very
dangerous. They were included in Burroughs Extended ALGOL to give access
to the B5000's Character Mode instructions, which had been a (somewhat
panicked) late addition to the architecture to support the large market
that Burroughs had in commercial business applications, and to support
the then very-new language COBOL. Bypassing hardware-enforced memory
protection was the price for that addition.

The B6500/6700/7700 did indeed fix the lack of memory protection with
the B5000/5500s Character Mode. It also fixed what I think was a more
serious ALGOL-related problem -- the inability of the B5000/5500 to
address intermediate nested scopes in a program. The older machine could
only address the global (outer block) variables and those in the current
local procedure. The hardware simply didn't have the ability to address
any nesting levels between those two.
>>>
>>> If Burroughs Large Systems -- or their Unisys successors -- had
>>> attracted more users, especially at universities, there might
>>> have been a reason to design a practical and portable version of
>>> ALGOL.
>>
>> Given that the original Algol didn't specify any I/O, so each
>> implementation did their own, it was hard to be portable. In fact
>> one "definition" of Pascal, was Algol plus I/O.
>
> Burroughs Extended ALGOL borrowed formatting from FORTRAN, changing
> FORMAT statements to FORMAT declarations. It seemed natural and
> obvious at the time; it would be surprising -- and unfortunate -- if
> no other ALGOL implementations did something similar.

Actually, the B5000/5500's formatting came from 220 BALGOL and was no
doubt heavily influenced by FORTRAN, although FORMAT declarations in
B5000 Extended ALGOL were certainly made more FORTRAN-like than they
were in BALGOL.

In the original B5000 ALGOL implementation, formatted I/O was the only
kind available. There was an ability to do unformatted reads and writes
via the RELEASE statement, but that manipulated physical buffers, much
like Direct I/O does on the later systems, and required the use of
Stream Procedures to access the data. The "array-row I/O" we all know
didn't arrive until the MCP was rewritten (in a higher-level language,
ESPOL) for the B5500.
>
>>
>>
>>> Burroughs Extended ALGOL and Large Systems' architecture were
>>> said to have been designed together; I attended a few Burroughs
>>> users' conferences until about 1982, and portability of source
>>> programs to other systems was a concept I never heard discussed.
>>>
>>> I would guess that most users of Burroughs Extended ALGOL had no
>>> idea that the language had a history outside Burroughs and that
>>> things called ALGOL-58 and ALGOL-60 existed. I certainly didn't
>>> have a clue.
>>
>> There were Algol compilers for many different architectures. For
>> example, the Univac 1100 series actually had two different ones.
>> Lack of compilers wasn't a major factor in Algol's lack of
>> popularity.

Let's not forget that there were two ALGOL compilers for the B5500 as
well, the original Extended ALGOL with Stream Procedures (plus several
other extensions with which you could undo both yourself and the system
if used improperly), and Compatible ALGOL (or XALGOL), which was
designed to be much safer to use, especially with timesharing systems,
and which would be much easier to port to the upcoming B6500. It used
the syntax that had been developed to support the B6500's string
operators, but was implemented for the B5500 by means of intrinsic
functions.

Let's also remember that the B5500 had two COBOL compilers -- the
original COBOL-61 and the later ANSI COBOL-68 compiler. So Univac's sins
in that regard, as mentioned in a later post to this thread, were hardly
unique.
>
> Burroughs users tended to be an insular bunch, and I was no
> exception. I recall one users conference attendee saying "If my firm
> changes systems, I'll change firms." I never imagined that I'd ever
> be working on anything else.
>
> <snip>
>
> Louis

Paul

-------- Original Message --------
Subject: Re: "In Defense of ALGOL"
From: Peter Flass <peter_flass@yahoo.com>
To:
Date: Fri Sep 02 2022 15:25:08 GMT-0700 (Pacific Daylight Time)

> Thomas Koenig <tkoenig@netcologne.de> wrote:
>> Louis Krupp <lkrupp@invalid.pssw.com.invalid> schrieb:
>>> See the second column on the first page:
>>>
>>> https://dl.acm.org/doi/pdf/10.1145/365230.365233
>>
>> Lack of standardized I/O was a huge design flaw, each compiler
>> had to roll its own, so there was no portability across systems.
>
> On the other hand, in reality simple I/O is pretty common across systems:
> read a line/write a line, etc. It’s only the more exotic operations that
> differ. If I were writing an ALGOL compiler I’d look at what the other guys
> did and imitate it, as compatibly as possible.

Which is pretty much what Burroughs did with the B5000/5500. Admittedly
it was FORTRAN-like, but that was the market they had to compete in. If
there ever was a "market" for ALGOL implementations, it certainly didn't
exist in the U.S. in the early 1960s.
>
>>
>> Kernighan wrote this in "UNIX: A History and a Memoir" that there
>> are three possibilities of doing I/O in a langue: Integrated into
>> the language itself (Fortran, Pascal), implemented in a library (C)
>> or not specifyling this at all. The last he called (from memory)
>> the least desirable option.
>>
>> And FORTRAN was not only groundbreaking in optimization and writing
>> formulas in a "natural" way. Its I/O was also radical and new,
>> the FORMAT statement was a big innovation

Paul

On 2022-09-03, Paul Kimpel <paul.kimpel@digm.com> wrote:

> Let's also remember that the B5500 had two COBOL compilers -- the
> original COBOL-61 and the later ANSI COBOL-68 compiler. So Univac's sins
> in that regard, as mentioned in a later post to this thread, were hardly
> unique.

Nor were Univac's "sins" unique to their 1100 series. The 90/30
(whose non-privileged instruction set was a clone of the IBM 360/50)
had three COBOL compilers: Basic COBOL, Extended COBOL, and COBOL-74.

In either case, though, I don't consider it a sin to release
a compiler that supports a later version of a language, while
retaining the old one for compatibility.

--
/~\ Charlie Gibbs | Microsoft is a dictatorship.
\ / <cgibbs@kltpzyxm.invalid> | Apple is a cult.
X I'm really at ac.dekanfrus | Linux is anarchy.
/ \ if you read it the right way. | Pick your poison.

Robin Vowels <robin.vowels@gmail.com> writes:

> PL/C was a pretty good implementation, and fast .
> Other non-IBM PL/I compilers included those for
> CDC, Univac, Burroughs, and on micros DR PL/I, Q1/Lite PL/I.

I'm surprised that no one has mentioned the GE/Honeywell PL/I compiler which
was used as the implementation language for much of Multics...

--
Rich Alderson news@alderson.users.panix.com
Audendum est, et veritas investiganda; quam etiamsi non assequamur,
omnino tamen proprius, quam nunc sumus, ad eam perveniemus.
--Galen

According to Rich Alderson <news@alderson.users.panix.com>:
>Robin Vowels <robin.vowels@gmail.com> writes:
>
>
>> PL/C was a pretty good implementation, and fast .
>> Other non-IBM PL/I compilers included those for
>> CDC, Univac, Burroughs, and on micros DR PL/I, Q1/Lite PL/I.
>
>I'm surprised that no one has mentioned the GE/Honeywell PL/I compiler which
>was used as the implementation language for much of Multics...

That was the Frebourghouse compiler.

Its front end was used for variety of other PL/I compilers such as the DEC VAX
one described in Engineering a Compiler

https://openlibrary.org/books/OL3489906M/Engineering_a_compiler

--
Regards,
John Levine, johnl@taugh.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. https://jl.ly