Re: iBM System/3 FORTRAN for engineering/science work? [message #411731 is a reply to message #411723] |
Thu, 07 October 2021 02:18 |
|
Originally posted by: Bob Eager
On Wed, 06 Oct 2021 17:40:29 -0700, Quadibloc wrote:
> On Wednesday, October 6, 2021 at 6:49:03 AM UTC-6, Bob Eager wrote:
>
>> And both L and LOWL versions are on the site.
>>
>> http://www.ml1.org.uk/implementation.html
>
> As you may have seen from my later posts, I eventually did find them
> myself; between my posts and yours, Waldek Hebisch, who originally asked
> about C and LOWL sources,
> should be able to find what he needs.
>
> John Savard
Yes, I thought I'd highlight the fact that the L version (even if only of
historical interest) is there too.
THere's also a BCPL version somewhere, but not on the site.
--
Using UNIX since v6 (1975)...
Use the BIG mirror service in the UK:
http://www.mirrorservice.org
|
|
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412076 is a reply to message #412073] |
Tue, 02 November 2021 08:07 |
Quadibloc
Messages: 4399 Registered: June 2012
Karma: 0
|
Senior Member |
|
|
On Tuesday, November 2, 2021 at 1:18:57 AM UTC-6, si...@situ.com wrote:
> On 2021-07-03, John Levine <jo...@taugh.com> wrote:
>>
>> Um, sendmail.cf uses m4, which is why it's opaque.
>
> I must confess that i still edit sendmail.cf by hand.
>
> Sometimes.
I did a Google search to get more information about this,
and the first result I got claimed that
sendmail.mc
is written in m4, and is what should be edited by hand.
Instead,
sendmail.cf
is what resdults after sendmail.mc ils compiled using
the m4 compiler.
John Savard
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412077 is a reply to message #409501] |
Tue, 02 November 2021 10:54 |
|
Originally posted by: Grant Taylor
On 6/30/21 11:12 PM, Thomas Koenig wrote:
> Seems far more readable than m4 (which is the serious programming
> language that I would consier most unreadable).
I learned about m4 as part of my Sendmail administration.
I've since adopted m4 to make multiple instances of -- what I describe
as -- an object oriented configuration language for things like:
1) the OpenSSH /client/ configuration file,
2) RWhoIs data
router(ip(...) ip(...) name(...))
router(name(...) ip(...) ip(...) ip(...))
router(comment(...) name(...))
3) the /etc/hosts file
I also use m4 for a number of other things.
I find the general purpose macro language to be quite useful.
--
Grant. . . .
unix || die
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412078 is a reply to message #412076] |
Tue, 02 November 2021 11:28 |
cross
Messages: 11 Registered: May 2013
Karma: 0
|
Junior Member |
|
|
In article <8c62a625-f704-4ca8-ac9c-ef17e8952f02n@googlegroups.com>,
Quadibloc <jsavard@ecn.ab.ca> wrote:
> On Tuesday, November 2, 2021 at 1:18:57 AM UTC-6, si...@situ.com wrote:
>> On 2021-07-03, John Levine <jo...@taugh.com> wrote:
>>>
>>> Um, sendmail.cf uses m4, which is why it's opaque.
>>
>> I must confess that i still edit sendmail.cf by hand.
>>
>> Sometimes.
>
> I did a Google search to get more information about this,
> and the first result I got claimed that
>
> sendmail.mc
>
> is written in m4, and is what should be edited by hand.
> Instead,
>
> sendmail.cf
>
> is what resdults after sendmail.mc ils compiled using
> the m4 compiler.
The sendmail configuration language is its own special
thing, based on production rewriting systems. Legend has
it that Eric Allman had taken a class on this and thought
it would be a neat hack to apply it to rewriting email
headers, which at the time was a big problem because of
the variety of email standards and bridges between them.
It was known to be nearly impossibly opaque, so eventually,
a set of macros written in M4 was created to _generate_
`sendmail.cf`. However, some folks continued to edit the
generated file by hand, and presumably still do.
I don't really try to run email systems anymore: spam
has made it require way too much effort for too little
gain, but for a long time sendmail was just the way most
mail was moved over the Internet. Then things like the
IDA toolkit for it popped up, and then qmail and Postfix
came along and some others were floating around (exim,
zmailer, MMDF) and it was time to move on. Still, because
of the generality of the sendmail language, it continued
to move lots of email, and presumably still does in some
places.
- Dan C.
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412344 is a reply to message #412076] |
Mon, 15 November 2021 20:50 |
sidd
Messages: 239 Registered: July 2012
Karma: 0
|
Senior Member |
|
|
On 2021-11-02, Quadibloc <jsavard@ecn.ab.ca> wrote:
> On Tuesday, November 2, 2021 at 1:18:57 AM UTC-6, si...@situ.com wrote:
>> On 2021-07-03, John Levine <jo...@taugh.com> wrote:
>>>
>>> Um, sendmail.cf uses m4, which is why it's opaque.
>>
>> I must confess that i still edit sendmail.cf by hand.
>>
>> Sometimes.
>
> I did a Google search to get more information about this,
> and the first result I got claimed that
>
> sendmail.mc
>
> is written in m4, and is what should be edited by hand.
Yes, i know. Nevertheless, i do. But only sometimes.
sidd
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412354 is a reply to message #409020] |
Wed, 17 November 2021 12:47 |
Quadibloc
Messages: 4399 Registered: June 2012
Karma: 0
|
Senior Member |
|
|
I was just looking through some old issues of Datamation for something else, and I saw an ad for PI*FORT.
This was claimed to be a "more accurate" Fortran than IBM's for the System/3.
Why? It did its arithmetic in decimal, so results wouldn't look funny!
John Savard
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412355 is a reply to message #412354] |
Wed, 17 November 2021 13:06 |
scott
Messages: 4237 Registered: February 2012
Karma: 0
|
Senior Member |
|
|
Quadibloc <jsavard@ecn.ab.ca> writes:
> I was just looking through some old issues of Datamation for something else, and I saw an ad for PI*FORT.
> This was claimed to be a "more accurate" Fortran than IBM's for the System/3.
> Why? It did its arithmetic in decimal, so results wouldn't look funny!
>
One might point out that decimal floating point is inherently more
_precise_ than binary floating point, and thus, likely more _accurate_
in certain boundary cases.
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412356 is a reply to message #412354] |
Wed, 17 November 2021 13:06 |
Dan Espen
Messages: 3867 Registered: January 2012
Karma: 0
|
Senior Member |
|
|
Quadibloc <jsavard@ecn.ab.ca> writes:
> I was just looking through some old issues of Datamation for something else, and I saw an ad for PI*FORT.
> This was claimed to be a "more accurate" Fortran than IBM's for the System/3.
> Why? It did its arithmetic in decimal, so results wouldn't look funny!
>
> John Savard
Also did arithmetic in binary.
--
Dan Espen
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412359 is a reply to message #412355] |
Wed, 17 November 2021 17:28 |
|
Originally posted by: Thomas Koenig
Scott Lurndal <scott@slp53.sl.home> schrieb:
> Quadibloc <jsavard@ecn.ab.ca> writes:
>> I was just looking through some old issues of Datamation for something else, and I saw an ad for PI*FORT.
>> This was claimed to be a "more accurate" Fortran than IBM's for the System/3.
>> Why? It did its arithmetic in decimal, so results wouldn't look funny!
>>
>
> One might point out that decimal floating point is inherently more
> _precise_ than binary floating point, and thus, likely more _accurate_
> in certain boundary cases.
One might point that out, but then one would have to state
the basis of that comparison.
If you base this on the number of digits, this is undoubtedly true -
a deciaml digit can carry much more information than a binary digit.
If you base it on the information that you can, for example,
store in a 64-bit word, then your claim needs some more explanation.
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412366 is a reply to message #412359] |
Wed, 17 November 2021 20:00 |
scott
Messages: 4237 Registered: February 2012
Karma: 0
|
Senior Member |
|
|
Thomas Koenig <tkoenig@netcologne.de> writes:
> Scott Lurndal <scott@slp53.sl.home> schrieb:
>> Quadibloc <jsavard@ecn.ab.ca> writes:
>>> I was just looking through some old issues of Datamation for something else, and I saw an ad for PI*FORT.
>>> This was claimed to be a "more accurate" Fortran than IBM's for the System/3.
>>> Why? It did its arithmetic in decimal, so results wouldn't look funny!
>>>
>>
>> One might point out that decimal floating point is inherently more
>> _precise_ than binary floating point, and thus, likely more _accurate_
>> in certain boundary cases.
>
> One might point that out, but then one would have to state
> the basis of that comparison.
>
> If you base this on the number of digits, this is undoubtedly true -
> a deciaml digit can carry much more information than a binary digit.
>
> If you base it on the information that you can, for example,
> store in a 64-bit word, then your claim needs some more explanation.
Binary floating point cannot precisely represent all valid
decimal floating point numbers.
|
|
|
Re: iBM System/3 FORTRAN for engineering/science work? [message #412369 is a reply to message #412366] |
Wed, 17 November 2021 20:51 |
John Levine
Messages: 1405 Registered: December 2011
Karma: 0
|
Senior Member |
|
|
According to Scott Lurndal <slp53@pacbell.net>:
> Thomas Koenig <tkoenig@netcologne.de> writes:
>> Scott Lurndal <scott@slp53.sl.home> schrieb:
>>> Quadibloc <jsavard@ecn.ab.ca> writes:
>>>> I was just looking through some old issues of Datamation for something else, and I saw an ad for PI*FORT.
>>>> This was claimed to be a "more accurate" Fortran than IBM's for the System/3.
>>>> Why? It did its arithmetic in decimal, so results wouldn't look funny!
>>>
>>> One might point out that decimal floating point is inherently more
>>> _precise_ than binary floating point, and thus, likely more _accurate_
>>> in certain boundary cases.
>>
>> One might point that out, but then one would have to state
>> the basis of that comparison. ...
> Binary floating point cannot precisely represent all valid
> decimal floating point numbers.
That is true, but it is irrelevant. With sufficient bits you can get accurate
results and rounding using binary arithmetic, and add 0.1 and 0.1 and get 0.2.
It was old news when I was taking numerical analysis classes in the 1970s.
It is certainly easier to do decimal rounding with a decimal representation but
that's not the same question.
--
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
|
|
|
|