[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
tom brough wrote: > > Someone told me recently that as little as 25% of the functionality of > an application is ever used. Is anyone aware of any references to > research in this area that I can follow up as I am reluctant to use > "word of mouth" statistics without some references to sources. The question is, I think, too vague to quantify - what is a feature has kept the software statisticians busy for a long time. Some have a vague definition based on function points - which might be what is referred to. For example if an OS ships with a Microsoft network driver and a Novell Netware one, if you don't use Novell's networking protocols in your network, is that not using functionality, or is that selecting between "equivalent" options? One could give a more objective answer for something like code coverage, but it wouldn't be especially helpful since it would depend greatly on coding style and language. i.e. it could have lots of code to handle obscure error conditions which is a good thing, but might not have many "features". It would still also count things like unused protocols. Similar OpenOffice Excel has a load of functions for accounting which I'll probably never use, but part of the utility of this kind of software is having those functions for the rare occasion you do need to calculate the cumulative principal on a loan to be paid between two periods. Otherwise one is left with anecdote. One of the best anecdotes was Microsoft Word, where it was claimed that the majority of new features requested at one point in life cycle were already in the product (and this was after it had got through Microsoft first few lines of defence - i.e. people who support and market Word had looked at the list and not noticed the features already exist). Might be worth tracking down the source of that one. GNU "ls" has about 56 command line arguments, and there are at least 26 I've never used in the 25 years or so I've been using it. I guess for command line junkies which options are used might generate some sort of objective measure. Although does it really matter if you've never done "ls --version" since the code is there for the one time you actually need to find this out. I'd have used a lot fewer if I'd code more script in Perl than Bash/Korn/Bourne over the years - does that make "ls" less good in some way? -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/linux_adm/list-faq.html