Some very interesting comments made by the speakers such as
- in one industrial case study presented about the use of pair programming, very few defects were found by the "navigator" of the pair.
- you can use rock-paper-scissors at the beginning of each day to determine who starts driving and who starts navigating.
- a Scrum case study (about 1000 KLOC) showed "linear or better" productivity increases by adding people. Note this is opposite to Brooks' "adding people to a later project makes it later". The general claim by the speaker was that by going to Scrum you can double productivity, in contrast to outsourcing which he claimed gives a 20 percent productivity increase.
Other impressions: There's actually been quite a bit of research done on pair programming, from a software development productivity point of view and from a cognitive approach. The industry people seem to be reinventing the wheel on some of these basic things. Not surprising since it feels good to reinvent a good idea like the wheel :)
Seriously though, I recommend that software engineers flip through Barry Boehm's slides from his ICSE 2006 keynote slides (ICSE is the big practitioner-academic yearly conference). He notes (on slide 7) that we are losing our history:
Median ICSE 2005 paper has no reference before 1984-85
77% have no references before 1980
In any case, everyone should look at his figure on slide 9.
