Corbett's approach is scientific - all the conclusions he reaches are eminently reasonable and he uses the statistics to reach those conclusions.

But in many other cases, be they in the free and open source world or in the proprietary software world, things work the other way. You first come to a conclusion and then try to obtain statistics to support that conclusion.

Anyone who has even the slightest inkling of how statistics can be used to manipulate conclusions will always be suspicious about surveys. I have seen umpteen surveys from the proprietary software crowd - which always tend to paint a positive picture of proprietary software. Not surprising at all.

The other side of the coin is the same. Surveys from proponents of free and open source software always mysteriously end up favouring the FOSS community and its software.

Not for nothing did the great Benjamin Disraeli once refer to statistics as being the third kind of lies - after lies and damn lies.

Recently I encountered a new twist on the use of statistics - where the opponent of one brand of software funds the collection of statistics about its rival. Yes, I'm talking about the misnamed Open Source Census being conducted by a company known as OpenLogic. (When companies add the word open here and there, it is meant to convey that things are above board - much in the same way that charlatans try to sell soap labelled "halal" in Muslim countries. Halal, incidentally, refers to a method of slaughter.)