Programs Without Source Code 
Author Message
 Programs Without Source Code


Quote:

>>(Anyway, competent programmers should not have created a Y2k problem in
>>the first place.)

>Careful with those asides, Craig. Can you say 'flamebait'?

>Once more, slowly and quietly. In the '60s and '70s there were good,
>pressing reasons for the six-digit date convention. By the 1980s, when
>young whippersnappers like me came into the industry, those reasons
>didn't apply - but by then it was a firmly-entrenched convention,
>which in itself makes it inappropriate to label everyone who stuck to
>it as incompetent. Even around 1990 I remember being regarded as a
>dangerous radical for insisting on eight-digit dates in my code (in
>defiance of project standards); it'd be ridiculous to conclude from
>that that I was the only competent programmer in the place. (I'm not
>saying it'd be wrong, just that it wouldn't be a valid deduction).

>Different people saw it coming at different times, that's all there is
>to it - Cory appears to have seen the light while I was still studying
>English Literature. We've all been stuck with the work of earlier
>programmers who hadn't seen the problem, or who were working early
>enough for the over-riding counter-arguments to apply.

It took a 2 by 4 in my case.  In 1969 I was a junior programmer doing
accounting applications. One of the old hands, I forget his name, told
me about a 'problem at the end of the century'.  I didn't really think
about it but I remember a bunch of us having a bull session.  I recall
him laughing about it.  I kept building non-Y2K compliant applications.

On December 1, 1979, I was the operating systems maintenance manager at
a time sharing service bureau.  All the machines in our multi-CPU
complex failed and I found the string '000197AF' in the dump.   I fixed
the code and got us back in business but not without a lot of stress and
visits from 'Executive Row'.  We were less than 30 days from bankruptcy.

It was bad code, it was our code, the code was developed by two
programmers who had left the company several months earlier.  Since my
department was sort of related to theirs, I was the stuckee.  Over the
next couple years, I invested in revising and improving their code.

There's lots of marginal code out there, code that will fail, not just
produce incorrect results, fail as in, 0Cx abends.   And you companies
out there, you better have lots of talent on the floor to fix these
problems.

Quote:
>And now, back to your scheduled slugfest.

>Phil

>--

>Editor, NEWS/400.uk             +44 (0)161 929 0777
>"Usenet is not a boat" - Angus Johnston

Cory Hamasaki       http://www.*-*-*.com/
HHResearch Co.      OS/2 Webstore & Newsletter
REDWOOD        


Sun, 12 Sep 1999 03:00:00 GMT  
 Programs Without Source Code

Quote:

> On December 1, 1979, I was the operating systems maintenance manager at
> a time sharing service bureau.  All the machines in our multi-CPU
> complex failed and I found the string '000197AF' in the dump.   I fixed
> the code and got us back in business but not without a lot of stress and
> visits from 'Executive Row'.  We were less than 30 days from bankruptcy.

On Monday, December 28, 1992, the beeper went off during my week of rotation.
Our most critical weekly job had abended.  We had three hours to fix the job,
and build a file to be transmitted to the printing contractor.  Under the
contractual arrangement, if the deadline was missed and it was our fault we
paid the contracting company something like a $50,000 penalty.

The file was delivered about 24 hours late and I was up all night.  And
yes, the exective row was awakened and wanted constant status updates.

The problem was traced to an assembler date routine.  The program name was
aliased and I had a hell of time locating the source code.  These days I know
COBOL much better than 370 assembler and it took a long time to determine
that the date routine had a very subtle error only on certain dates in leap
years.  The routine was used to generate a future date (in our case it had to
be a Monday, and was not), and naturally it only works with 2-digit years.

Besides being difficult to locate the source, and assembler talent to fix it,
there was a political problem in that neither technical services nor the
development departments could agree on who owned the problem.  Normally, TS
owns utility subprograms that are used by multiple departments.  This routine
had been developed years earlier by a different applications group.  It had
never been documented as a public utility, but had crept into several
systems.

It took over a week to solve the political problems, fix the code, test it,
and implement it.  That program is still in production, and will fail on any
date that falls after 12/31/99.  We've found it still in use in supposedly
Y2K compliant systems, and we have warned everyone working on Y2K conversions
to replace it with a newer utility that works.

I recently fixed a different assembler date routine, to window it for
1950 to 2049.  It has been in production at least 15 years, and I discovered
it was loaded with leap year errors prior to 1969.  Naturally, it's used in
so many places that it's too much work to replace it.  So I've tested it with
every date in the entire 100-year range.

The thought of multiple concurrent system failures as bad as that one makes
me break out in a cold sweat.  It should put the fear of God into anyone.

Quote:

> There's lots of marginal code out there, code that will fail, not just
> produce incorrect results, fail as in, 0Cx abends.   And you companies
> out there, you better have lots of talent on the floor to fix these
> problems.

> Cory Hamasaki       http://www.kiyoinc.com
> HHResearch Co.      OS/2 Webstore & Newsletter
> REDWOOD

Cory, I couldn't agree with you more!

Arnold Trembley
Software Engineer I (just a job title, still a programmer)
MasterCard International
St. Louis, Missouri



Mon, 13 Sep 1999 03:00:00 GMT  
 
 [ 3 post ] 

 Relevant Pages 

1. Programs Without Source Code

2. Determining links without source code ?

3. Translating fortran source code to matlab program (*.m) code

4. Deploy without distribute python source code

5. Data Recovery SOURCE CODE ( SOURCE CODES of Professional Data Recovery Software )

6. Converting FORTRAN source code to C source code

7. Best source code control program

8. Best source code control program

9. New FTP Archive for Source Code and Programming Utilities

10. Help needed: program that prints out its source code when RUN

11. Source Code Translation Programs ???

12. asm source code of modem accessing program?

 

 
Powered by phpBB® Forum Software