Prior to the 21st century, researchers often had access to data from only one or
two projects. This meant theories of software development were built from limited
data. But in the data-rich 21st century, researchers have access to all the data they
need to test the truisms of the past. And what they’ve found is most surprising:
• In stark contrast to the assumptions of much prior research, pre- and post-
release failures are not connected.25
• Static code analyzers perform no better than simple statistical predictors.26
• The language construct GOTO, as used in contemporary practice, is rarely
considered harmful.27
• Strongly typed languages are not associated with successful projects.28
• Developer beliefs are rarely backed by any empirical evidence.3
• Test-driven development is not any better than “test last.”29
• Delayed issues are not exponentially more expensive to fix.30
• Most “bad smells” should not be fixed.31,32
timm
commented
5 years ago
Prior to the 21st century, researchers often had access to data from only one or two projects. This meant theories of software development were built from limited data. But in the data-rich 21st century, researchers have access to all the data they need to test the truisms of the past. And what they’ve found is most surprising: • In stark contrast to the assumptions of much prior research, pre- and post- release failures are not connected.25 • Static code analyzers perform no better than simple statistical predictors.26 • The language construct GOTO, as used in contemporary practice, is rarely considered harmful.27 • Strongly typed languages are not associated with successful projects.28 • Developer beliefs are rarely backed by any empirical evidence.3 • Test-driven development is not any better than “test last.”29 • Delayed issues are not exponentially more expensive to fix.30 • Most “bad smells” should not be fixed.31,32