NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/software/5054

Internet Message Format | 1992-12-18 | 1.6 KB

Path: sparky!uunet!ogicse!cadreor!dbsoft!dws
From: [e-mail] (David W. Smith)
Newsgroups: comp.software-eng
Subject: Re: Software Inspections.  How many does it
Message-ID: <[e-mail]>
Date: 17 Dec 92 23:31:05 GMT
Article-I.D.: menlo.1992Dec17.233105.13228
References: <[e-mail]> <[e-mail]>
Organization: Cadre Technologies Inc., Menlo Park Operation
Lines: 23

In article <[e-mail]> mentions one of the downsides
of adding people to a review team.

>But mn wasn't talking about *independent* review teams.  He was talking about
>adding more people to a team.  In this case, I suggest that the amount/number
>of errors detected per person added to the team decreases rapidly for team sizes
>beyond four or five.

One cause that I see for the decrease in error detection in larger teams is
that people fall prey to the notion that their participation isn't really all
that important, and that the rest of the team will cover for them if they
aren't prepared.  To give the illusion of being prepared, they pick then nits
with inconsequential matters, such as grammar and spelling in comments, and
the review slows way down.

It helps here to assign specific roles.  For example, if I'm tasked with
reviewing code for memory leaks, I'm more likely to believe that the use of
my time is valuable (and less likely to hope that the team will cover for me)
and will make a greater effort to block out time to carefully review the code.

--
Dave Smith, Cadre Technologies Inc. - Database Tools Division
4500 Bohannon Drive, Suite 150, Menlo Park, CA  94025