The top ten reasons embedded projects get into trouble (2002)
Embedded systems are notoriously expensive and time-consuming parts of modern technology. According to Norman Augustine, former CEO of Lockheed, they became a necessity for defense contractors in the late 70s due to increased business demands despite physics limitations. Today, high-performance aircraft firmware alone consumes about half of the total price of the plane.
The issue seems to be getting worse as embedded software doubles in size every 10 months. This size growth, coupled with the exponential relationship between schedule and size, suggests that within a few years, firmware development could consume nearly the entire world's GDP.
The situation is worsened by the fact that software people often struggle to secure adequate funding for essential tools like static analyzers. The belief that porting GNU is sufficient is an early warning sign of potential disaster. Agile methods have influenced software development, but in the embedded world, careful analysis is crucial as design decisions are often rigid and costly to change later.
Poorly structured code can lead to failure in meeting real-time requirements. Choosing the wrong CPU or insufficient memory can also cause project failure from the start.
Some argue that languages like Ada and SPARK, which have fewer bugs than C, are better choices. Ada, in particular, is favored for safety-critical applications. However, C is still commonly used and can be effective with disciplined use and proper tools. The misuse of C is considered a bad science issue, either due to poor analysis of real-world events or insufficient A/D resolution, precision, or accuracy in measurements.
The second type of issue arises from uncovering new or unknown scientific information, but in most cases, the science is known and needs proper application.
Rewritten by urgent.news; may contain errors — check the original.