This chapter is from the book
This chapter is from the book
This chapter is from the book
1.2 The Inherent Complexity of Software
A dying star on the verge of collapse, a child learning how to read, white blood cells rushing to attack a virus: These are but a few of the objects in the physical world that involve truly awesome complexity. Software may also involve elements of great complexity; however, the complexity we find here is of a fundamentally different kind. As Brooks points out, "Einstein argued that there must be simplified explanations of nature, because God is not capricious or arbitrary. No such faith comforts the software engineer. Much of the complexity that he must master is arbitrary complexity" [1].
Defining Software Complexity
We do realize that some software systems are not complex. These are the largely forgettable applications that are specified, constructed, maintained, and used by the same person, usually the amateur programmer or the professional developer working in isolation. This is not to say that all such systems are crude and inelegant, nor do we mean to belittle their creators. Such systems tend to have a very limited purpose and a very short life span. We can afford to throw them away and replace them with entirely new software rather than attempt to reuse them, repair them, or extend their functionality. Such applications are generally more tedious than difficult to develop; consequently, learning how to design them does not interest us.
Instead, we are much more interested in the challenges of developing what we will call industrial-strength software. Here we find applications that exhibit a very rich set of behaviors, as, for example, in reactive systems that drive or are driven by events in the physical world, and for which time and space are scarce resources; applications that maintain the integrity of hundreds of thousands of records of information while allowing concurrent updates and queries; and systems for the command and control of real-world entities, such as the routing of air or railway traffic. Software systems such as these tend to have a long life span, and over time, many users come to depend on their proper functioning. In the world of industrial-strength software, we also find frameworks that simplify the creation of domain-specific applications, and programs that mimic some aspect of human intelligence. Although such applications are generally products of research and development, they are no less complex, for they are the means and artifacts of incremental and exploratory development.
The distinguishing characteristic of industrial-strength software is that it is intensely difficult, if not impossible, for the individual developer to comprehend all the subtleties of its design. Stated in blunt terms, the complexity of such systems exceeds the human intellectual capacity. Alas, this complexity we speak of seems to be an essential property of all large software systems. By essential we mean that we may master this complexity, but we can never make it go away.
Why Software Is Inherently Complex
As Brooks suggests, "The complexity of software is an essential property, not an accidental one" [3]. We observe that this inherent complexity derives from four elements: the complexity of the problem domain, the difficulty of managing the development process, the flexibility possible through software, and the problems of characterizing the behavior of discrete systems.
The Complexity of the Problem Domain
The problems we try to solve in software often involve elements of inescapable complexity, in which we find a myriad of competing, perhaps even contradictory, requirements. Consider the requirements for the electronic system of a multiengine aircraft, a cellular phone switching system, or an autonomous robot. The raw functionality of such systems is difficult enough to comprehend, but now add all of the (often implicit) nonfunctional requirements such as usability, performance, cost, survivability, and reliability. This unrestrained external complexity is what causes the arbitrary complexity about which Brooks writes.
This external complexity usually springs from the "communication gap" that exists between the users of a system and its developers: Users generally find it very hard to give precise expression to their needs in a form that developers can understand. In some cases, users may have only vague ideas of what they want in a software system. This is not so much the fault of either the users or the developers of a system; rather, it occurs because each group generally lacks expertise in the domain of the other. Users and developers have different perspectives on the nature of the problem and make different assumptions regarding the nature of the solution. Actually, even if users had perfect knowledge of their needs, we currently have few instruments for precisely capturing these requirements. The common way to express requirements is with large volumes of text, occasionally accompanied by a few drawings. Such documents are difficult to comprehend, are open to varying interpretations, and too often contain elements that are designs rather than essential requirements.
A further complication is that the requirements of a software system often change during its development, largely because the very existence of a software development project alters the rules of the problem. Seeing early products, such as design documents and prototypes, and then using a system once it is installed and operational are forcing functions that lead users to better understand and articulate their real needs. At the same time, this process helps developers master the problem domain, enabling them to ask better questions that illuminate the dark corners of a system's desired behavior.
)
Because a large software system is a capital investment, we cannot afford to scrap an existing system every time its requirements change. Planned or not, systems tend to evolve over time, a condition that is often incorrectly labeled software maintenance. To be more precise, it is maintenance when we correct errors; it is evolution when we respond to changing requirements; it is preservation when we continue to use extraordinary means to keep an ancient and decaying piece of software in operation. Unfortunately, reality suggests that an inordinate percentage of software development resources are spent on software preservation.
The Difficulty of Managing the Development Process
The fundamental task of the software development team is to engineer the illusion of simplicity—to shield users from this vast and often arbitrary external complexity. Certainly, size is no great virtue in a software system. We strive to write less code by inventing clever and powerful mechanisms that give us this illusion of simplicity, as well as by reusing frameworks of existing designs and code. However, the sheer volume of a system's requirements is sometimes inescapable and forces us either to write a large amount of new software or to reuse existing software in novel ways. Just a few decades ago, assembly language programs of only a few thousand lines of code stressed the limits of our software engineering abilities. Today, it is not unusual to find delivered systems whose size is measured in hundreds of thousands or even millions of lines of code (and all of that in a high-order programming language, as well). No one person can ever understand such a system completely. Even if we decompose our implementation in meaningful ways, we still end up with hundreds and sometimes thousands of separate modules. This amount of work demands that we use a team of developers, and ideally we use as small a team as possible. However, no matter what its size, there are always significant challenges associated with team development. Having more developers means more complex communication and hence more difficult coordination, particularly if the team is geographically dispersed, as is often the case. With a team of developers, the key management challenge is always to maintain a unity and integrity of design.
The Flexibility Possible through Software
A home-building company generally does not operate its own tree farm from which to harvest trees for lumber; it is highly unusual for a construction firm to build an onsite steel mill to forge custom girders for a new building. Yet in the software industry such practice is common. Software offers the ultimate flexibility, so it is possible for a developer to express almost any kind of abstraction. This flexibility turns out to be an incredibly seductive property, however, because it also forces the developer to craft virtually all the primitive building blocks on which these higher-level abstractions stand. While the construction industry has uniform building codes and standards for the quality of raw materials, few such standards exist in the software industry. As a result, software development remains a labor-intensive business.
The Problems of Characterizing the Behavior of Discrete Systems
If we toss a ball into the air, we can reliably predict its path because we know that under normal conditions, certain laws of physics apply. We would be very surprised if just because we threw the ball a little harder, halfway through its flight it suddenly stopped and shot straight up into the air.1 In a not-quite-debugged software simulation of this ball's motion, exactly that kind of behavior can easily occur.
Within a large application, there may be hundreds or even thousands of variables as well as more than one thread of control. The entire collection of these variables, their current values, and the current address and calling stack of each process within the system constitute the present state of the application. Because we execute our software on digital computers, we have a system with discrete states. By contrast, analog systems such as the motion of the tossed ball are continuous systems. Parnas suggests, "when we say that a system is described by a continuous function, we are saying that it can contain no hidden surprises. Small changes in inputs will always cause correspondingly small changes in outputs" [4]. On the other hand, discrete systems by their very nature have a finite number of possible states; in large systems, there is a combinatorial explosion that makes this number very large. We try to design our systems with a separation of concerns, so that the behavior in one part of a system has minimal impact on the behavior in another. However, the fact remains that the phase transitions among discrete states cannot be modeled by continuous functions. Each event external to a software system has the potential of placing that system in a new state, and furthermore, the mapping from state to state is not always deterministic. In the worst circumstances, an external event may corrupt the state of a system because its designers failed to take into account certain interactions among events. When a ship's propulsion system fails due to a mathematical overflow, which in turn was caused by someone entering bad data in a maintenance system (a real incident), we understand the seriousness of this issue. There has been a dramatic rise in software-related system failures in subway systems, automobiles, satellites, air traffic control systems, inventory systems, and so forth. In continuous systems this kind of behavior would be unlikely, but in discrete systems all external events can affect any part of the system's internal state. Certainly, this is the primary motivation for vigorous testing of our systems, but for all except the most trivial systems, exhaustive testing is impossible. Since we have neither the mathematical tools nor the intellectual capacity to model the complete behavior of large discrete systems, we must be content with acceptable levels of confidence regarding their correctness.