You are on page 1of 2

INTRODUCTION

SOFTWARE development organizations face the difficult problem of producing high-qualit ! lo"-defect soft"are on time and on-budget# A $S government stud %&' estimated that soft"are defects are costing the $S econom appro(imatel )*+#* billion per ear# Several high-profile soft"are failures have helped to focus attention on this problem! including the failure of the ,os Angeles airport-s air traffic control s stem in .//0! the 1ortheast po"er blac2out in .//3 4"ith an estimated cost of )5-)&/ billion6! and t"o failed 1ASA 7ars missions in &+++ and ./// 4"ith a combined cost of )3./ million6# 8arious approaches to addressing this epidemic of budget and schedule overruns and soft"are defects have been proposed in the academic literature and applied in soft"are development practice# T"o of these approaches are soft"are inspection and test-driven development 4T996# :oth have advantages and disadvantages! and both are capable of reducing soft"are defects %.'! %3'! %0'! %*'# Soft"are inspection has been the focus of more than 0// academic research papers since its introduction b Fagan in &+;5 %5'# Soft"are inspection is a formal method of inspecting soft"are artifacts to identif defects# This method has been in use for more than 3/ ears and has been found to be ver effective at reducing soft"are defects# Fagan reported soft"are defect reduction rates bet"een 55 and <. percent %5'# While an soft"are development artifact ma be inspected! most of the soft"are inspection literature deals "ith the inspection of program code# =n this stud ! "e limit inspections to program code! and "e refer to these inspections as code inspections# T99 is a relativel ne" soft"are development practice in "hich unit tests are "ritten before program code# 1e" tests are "ritten before features are added or changed! and ne" features or changes are generall considered complete onl "hen the ne" tests and an previousl "ritten tests succeed# T99 usuall involves the use of unit-testing frame"or2s 4such as >$nit& for >ava development6 to support the development of automated unit tests and to allo" tests to be e(ecuted frequentl as ne" features or modifications are introduced# Although results have been mi(ed! some research has sho"n that T99 can reduce soft"are defects b bet"een &< and */ percent %.'! %3'! "ith one stud sho"ing a reduction of up to +& percent %;'! "ith the added benefit of eliminating defects at an earlier stage of development than code inspection# T99 is normall described as a method of soft"are design! and as such! has benefits that go be ond testing and defect reduction# ?o"ever! in this stud "e limit our anal sis to a comparison of the defect reduction benefits of the methods and do not consider other benefits of either approach# E(isting research does not sufficientl assess "hether T99 is a useful supplement or a viable alternative to code inspection for purposes of reducing soft"are defects# @revious research has compared the defect reduction benefits of code inspection and soft"are testingAmuch of "hich is summarized b Runeson et al# %<'# ?o"ever! the current high adoption rates of T99 indicate the timeliness and value of specific comparisons of code inspection and T99# The focus of this

stud is a comparison of the defect rates and relative costs of these t"o methods of soft"are defect reduction# =n the stud ! "e see2 to ans"er the follo"ing t"o main questionsB &# Which of these t"o methods is most effective at reducing soft"are defectsC .# What are the relative costs of these soft"are defect reduction methodsC The soft"are inspection literature t picall uses the term DdefectE to mean Dfault#E This literature contains a "ellestablished practice of categorizing defects as either DmaForE or DminorE %+'! %&/'# =n this paper! "e compare the effectiveness of code inspection and T99 in removing DmaFor defects#E The remainder of this paper is organized as follo"sB The ne(t section discusses the relevant research related to this stud # Section 3 describes the purpose and research questions# Section 0 describes the e(periment and the procedures follo"ed# Sections * and 5 describe the results and implications of the findings! and Section ; concludes "ith a discussion of the stud -s contributions and limitations#

You might also like