Good Ol' Bean Counter, where art thou:

There is this study published few years ago by Cap Gemini Ernst & Young which states the following: 65% of the in-memory image of a software program belongs to software developed by a third party.

Put this figure together with the NIST stats (80% of a project is testing and debugging) and the magical number is 52%. Any software project could use up to 52% of its resources (time, people, etc) executing somebody else's software.

And doing the marketing research update for the BP, I ran into Mercury's QTP 8.X.

This tool allows users to continue using capture & replay through another layer of indirection (their ActiveScreen feature)

So a capture & replay user just interested in:

button1_clicked(because is the only logic that she/he can modify)

has to wait for the tool to execute who knows how many screen repaints, OS and specific tool calls (all of which are third party) to get to the real test target:

the fabled "button1_clicked".

How many more tests could be executed instead? Aren't we always short of time for quality? Are we using the time we have wisely?

On the same token, if we could get rid of the extra layers of indirection. Could we test everything all the time?

The Mythical Man Month considered this not economically/physically feasible. But back then the computing power was in its infancy and most of the production of software was done manually. Is it possible now?