I was attending an offsite training, so I missed the presentation Ivar Jacobson given to our company, but had the pleasure to had dinner with him. Here was what I caught (might have misunderstanding):

You can do MDD, but forget MDA. MDA is not mature and only PHD student can understand and use it.

Use case vs User Story: Use case can be used to derive test cases, but user story can't.

Standard persistent format for UML

It is in the first version of UML, but every tool created its own format.

UML library: Rational has created some common asset management pieces, but not very useful.

UML 2.0 is too heavy. I was thinking of using essential UML (20% features) in development.

Today I had a chance to ask questions to Craig Larman and here was what I got.

Q: Do I need to create WBS before Iteration 1?

A: You spend 2~4 days in creating a backlog (feature list) before iteration 1, and spend 5% time of each iteration to do plan for current iteration, and team members create their own WBS in each iteration.

Q: Do I estimate size of project according to backlog?

A: Yes.

Some notes I caught (might be wrong):

Don't do resource pool, a stable team is better.

Microsoft did daily build 20 years ago, if you don't do, you are 20 years out of date.

It is a bad idea to "code a little, test a little". It is a good idea to "test a little, code a little".

User story can be documented as: "I want <a feature> because it has <this benefit>".

Abandon MS word document as requirement Spec, use wiki instead.

There are a lot of books proclaim that "Do not program prematurely". Many readers understand it as "Do not program" before you have a mature model. This is misleading because for Non-functional requirements, it is difficult to model without investigation. You should start investigating every new technology that are going to be used in your design, otherwise they won't work as your expected. New technology will always surprises you if you are not familiar to it. As an example, people prematurely think Multi-CPU will run faster than single CPU, but it is not always true; for some programs that do not take advantage of extra CPU resources, it can be SLOWER due to extra synchronization cost between CPUs. Windows service is part of our new system, then I spent a few days to familiar to service project in VC++ 9.0(which is a lot different from VC++ 6.0 version). Tcp will play an important role in the new system, then I spent efforts in playing with boost.asio. Don't postpone investigation to construction phase, otherwise you can't get right estimation for corresponding task, and also at the risk of blocking others from system integration.

MS MPI vs. MPICH2

Download MS MPI here

MPI vs OpenMP

PVM FAQ

TSP stated that "New and inexperienced developers are often reluctant to take on unfamiliar challenges. More experienced developers are eager to learn new methods and techniques."
This is not true. New members usually does not reject tough assignment in order to acquire recognition. More experienced developers usually familiar to the project and understand the risk and are more active to say no. Some experienced developers stopped learning and focus on personal investment.

The famous Maslow's Hierachy of Needs is cited in TSP. TSP claimed that self-actualization level is required for development work, which is not bulletproof in my opinion.There are several facts beyond Maslow's theory: People may sacrifice their life for religion or belief, like Galileo. Another fact is that "Self-actualization" may not be aligned with business goal, mostly likely be a personal hobby like sightseeing, etc. If there is possibility, many people will retire early, thus get work done is not self-actualization, I would emphasize on cultivating motivation in belonging and recognition level. When working as a team, we have the feeling of belonging and recognition in common, while self actualization is diverse.

Creative people needs for job security.
Agree, quite a lot people don't change job for better offer. They would prefer work for large company to small company. Most development professionals want to work on a cohesive and cooperative team. Many people only go to head hunter when they are unhappy to current environment, not current salary package.
The team members like to be successful.
Many team members want to do technical work what is interesting and that satisfies their personal goals and aspirations. There are also quite a lot of people like to do nonchallenging work, they are not akin to learning new knowledge.

I realized that TDD is not enough for quality ensurance one year ago -- unit test code was written but a use case still failed, since TDD does not focus at use case level, so I think an independent test team is important for high level validation&verification. Scott W. Ambler published a lengthy article on Feb 2008 issue of Dr. Dobb's, he also considered Agile Modeling in addition to test team.

Applied Software Project Management challenges the rule of "testing should always take exactly one-third (or some other fixed ration) of the programming time", which is promoted by The Art of Project Management. I'd support the latter since we are making guess and it is simpler to guess one factor instead of two.

Wideband Delphi Estimation process:

Choosing a qualified team
The team generates a task list of 10-20 major tasks and assumptions (WBS) based on vision and scope document in Kickoff meeting
Indivisual preparation offline estimation
Estimation session
Additional session to solve disagreement