Pelo
fato da qualidade não possuir uma única dimensão, atributo, ou característica,
a noção de qualidade normalmente é capturada em uma modelo que descreve as
características compostas e as relações entre elas. Para definir um modelo
de qualidade, deve haver uma clara distinção entre atributos internos e externos.
Qualidade pode ser medida de diversas formas - métricas e critérios devem
ser estabelecidos para satisfazer as necessidades do projeto de software.
Este artigo discute como aplicar métricas de software para melhorar a qualidade
e a confiabilidade de produtos de software, baseado nas melhores práticas
de desenvolvimento de software, práticas comercialmente aprovadas.
Abstract:
Software engineering has a reputation for poor quality, late delivery and
budget overrun. On past performance this reputation is well deserved. One
of the most significant problems is that it errors tend not to be detected
until it is too late to correct them within deadline and cost constraints.
The elimination of all errors is, at least in current industrial environments,
well beyond the capability of available software technology. Nevertheless
experience suggests that density of errors in released code can be halved
by in-process checks supported by suitable analytical tools. This does not
entail change of paradigms or languages. All it needs is the will to adopt
proven techniques.
Of all forms of software quality control, code inspection is by far the most
cost effective. When supported by good tools, the unit cost of static error
detection is typically one or two orders of magnitude less than detection
by dynamic methods. To make inspections work you need a coding standard setting
out requirements that inspected code must satisfy. This paper explains the
role of coding standards in code quality control.