loyal opposition
104 IEE E Softwar E Published by th e IEEE Computer Soc iety 0740-7459/0 9/$25. 00 © 2009 IE EE
In the last decade, the open source movement
has been a vital source of innovation affect-
ing soft ware development.
1
However, open
source community practices have provoked a
debate on software quality—namely, is open
sou rce sof twar e’s qual ity b etter than that of its
closed-source counterpart? Studies have attempted
to correlate metrics with software performance or
validate that metrics can actually predict software
systems’ fault proneness.
2
For open source soft-
ware, a strand of research has emerged on quanti-
fying software quality through code reading and
judging of open source codes.
3
Here, I attempt an
evidence-based or experimental software research
approach
4
to answer the question.
What is open source software?
Whereas you can define closed-source software
as a product created using traditional software
development methods, the definition of open
source software isn’t always straightforward.
This is because a software product can take at
least three paths to become open source. For ex-
ample, a collaborating open source community
developed the Linux kernel, an individual cre-
ated PGP (Pretty Good Privacy), and the Mozilla
browser was originally developed as proprietar y
software. One implication of this is that any con-
clusions about Linux might not hold true for all
open source products.
What Do We Know
from existing studies?
I evaluated studies focusing on performance and
quality rather than total cost of ownership. Mind-
craft conducted one such study, with participation
from RedHat and Microsoft engineers.
5
The engi-
neers tuned the tested systems to ensure maximum
performance. They performed tests on one- and
four-processor configu rations of Dell servers in-
stalled with Windows NT Server 4.0 and the Red
Hat Linux 6.0 kernel. The results showed that Win-
dows NT Server had faster t hroughput than Linux
on both configurations.
In 1999, Bloor Research experimentally com-
pared the maturity of Windows NT to that of
GNU/ Linux for one year.
6
The Windows NT sys-
tem suffered 68 failures, of which 26 were memory
related and eight were file system faults. The Linux
system experienced only one software-related fail-
ure. Bloor rated the Linux servers as good as or bet-
ter tha n the Windows servers in five of the s even ar-
eas monitored. The study has an obvious limitation.
Namely, it’s unclear whether 25 of the 26 memory
failures were due to the Windows OS or to the poor
quality of the intervention to repair the first error.
Ioannis Samoladas and his colleagues found
that the code quality of successive versions of open
source software was equa l to or better than that of
closed-source products.
7
Security Innovation pit-
ted a Windows 2000 system against a SuSE Linux
Enterprise Server 8 for one year. The company con-
cluded that Windows Server environments seem to
man ifest low er risk of inst abilit y when subjected to
modifications.
8
Diomidis Spinellis used CScout, a
source code analyzer, to investigate the code qual-
ity of four operating systems: FreeBSD 6.1, Open
Solaris, Linux 2.6.18, and major por tions of the
Windows Research Kernel.
9
He investigated met-
rics for file organization, code structure, code style,
preprocessing, and data organization. He found
that the open source products’ code quality wasn’t
markedly higher than that of the closed-source
products.
A Vulnerability study
You can define vulnerability as the lack of adequate
security. Security is software’s abilit y to prevent
Adenekan (Nick) Dedeke
The evidence is not that clear- cut.
is Linux Better than
Windows software?
Continued on p. 103
Authorized licensed use limited to: UNIVERSIDADE FEDERAL DO PARA. Downloaded on April 21, 2009 at 12:45 from IEEE Xplore. Restrictions apply.
