DEV Community

Cover image for Bugs, errors and causes from an 1975 paper
Lucian Ghinda
Lucian Ghinda

Posted on • Originally published at allaboutcoding.ghinda.com

Bugs, errors and causes from an 1975 paper

Discover the timeless relevance of a 1975 paper that sheds light on the causes of bugs and errors in system programs.

In this article, we delve into Albert Endres' findings and explore how understanding the problem, effective communication, and domain knowledge play crucial roles in reducing errors.

These principles remain highly relevant in current software development practices, where collaboration, clear communication, and comprehensive knowledge of the project's domain could be crucial for the success of the product.

About the paper

In 1975, Albert Endres published a paper called "An Analysis of Errors and their Causes in System Programs".

In this paper, they analyze the errors detected in an operating system DOS/VS developed by IBM Laboratories in Germany.

The system was released in 1973. The paper analyses:

- 500 modules, with an average of 360 lines per module and 480 lines of comments, resulting in 190K instructions and 60K comments.

- A total of 740 problems were found, of which 432 were classified as program errors.

Areas where errors where found

And here are some of the conclusions from this paper:

If we want to reduce the number of errors, then we should focus not only on better programming techniques but also on the problem definition, and understanding domain knowledge:

And based on the paper, here are the categories of causes for errors or bugs:

1. Technological

2. Organisational

3. Historic

4. Group dynamic

5. Individual

6. Other

An image showing the category of bugs and their description

Some conclusions:

  1. When considering quality in a software product, we should consider all aspects, not only the tools and programming techniques we use.

  2. When joining a product or team, it is important to understand the history of the product & team.

  3. To become a better developer, one must learn more than coding/engineering techniques.

  4. When we think about a possible future of using AI to augment programming, we should consider that if this study remains true it is not enough to reduce the number of errors. As half of them are not caused by code.


Enjoyed this article?

Join my Short Ruby News newsletter for weekly Ruby updates. For more Ruby learning resources, visit rubyandrails.info

Top comments (0)