It's acceptance rate for technical papers is very healthy- around one in five.

The conference suffers from having too broad a focus. Nevertheless, anyone who is anyone seems to go there so it is a great place to network.

(But if I am feeling nasty, I say that ICSE is the international conference on manual SE. These folks know so little about automatic inference and sometimes I want to lock them all in a dungeon and not let them out till they understand theorem proving. But that's just me.)

That's smaller than ICSE 2001 at Toronto, which had 975.

Answer: society under-values software. There are seats in the President's cabinet for agricultural and housing. These are well-understood problems dating back several centuries. But there is no "secretary for software", even though software is currently 10% of the American GNP and we know so little about it.

Strangely, we under-value software even though our software is so awful. Don't believe me? Well, why don't you compare...

• How thrill led you are when your car comes back from the garage, all tuned up.
• How terrified you are when, just before take-off, the captain announces that your plane has a brand new operating system.

Think how reliant we are on bad software solutions. For example, the next generation of commercial aircraft will use DHCP servers and ethernet to connect their equipment. Passengers logging in from their seats will use the same servers. Imagine the scene:

• Your computer boots up then says "new hardware Boeing 747 found, shall I install?".
• While your happily surfing the web in mid-flight, the terrorist sitting next to you is cracking into the fuel supply of the plane.

Speed of damage of viruses is increasing:

• Melissa: 1999, $300M damage in 4 days, 150,000 systems
• ILoveYou: May 2000, $10B damage in 24 hours in 500,000 systems
• Code red1: 400,000 systems infected in 14 hours, several billions of dollars
• Saphire/slammer (most recent): 75,000 systems infected in 10 minutes, $1B damage.

Scary stuff, right? But worse, according to one keynote speaker, nearly 3/4 of the security flaws used by these viruses should have been caught with standard SE methods. For example, simple buffer overflow and input validation covers 67% of all security exploits.

But why am I worried? Sounds like a good reason to sell more IV&V services.

More seriously, the security keynote said that security is perhaps one of the best things to happen to software engineering in decades. Security means that quality is now longer non-optional.

Law & Rothermel's paper on "impact analysis" was all about finding out what is effected by a change to some part of a program. Used a European Space Agency example and CODESURFER as its code slicing tool.

Gail Murphy's (et.al.) three papers were all about how to cheaply augment source code with some higher-level understanding of that code.

For those of you excited or alarmed about Agile methods or Extreme Programming, the Abrahamsson et.al. paper is an excellent (and short) comparison of Agile/Extreme with seven other related approaches.

Barry et.al. offered an insightful analysis of common types of software evolution.

Robyn Lutz presented the post-anomaly ODC stuff that we (partially) funded.

Jon Whittle talked to his "Requirements to Code via UML" stuff.

In the tools demo section, the WVU architectural assessment tool built by Hany, Katerina, et.al. was presented.

Heard that the Intelligent Systems BAA got 700+ proposals. Gasp. I put four proposals into that one and, at 700 proposals, it will be a real turkey shoot.

Also heard that HDCP (High Dependency Computing Program) is having some issues: hard to get time with various projects for them to generate test beds. Fancy that: NASA projects reluctant to get into bed with research? Who would have thought?

Lots of evidence shown that women's presence in CS is wwwaaaaayyyy out of line with other technical fields (e.g. mathematics, engineering).

Worst still, based on a recent survey that asked high-school students about their university plans, enrollments in CS will further decrease at least in the near future.

What to do? Well..

• CS departments that have equivalent male/female retention rate always had same sex support- someone to go to when you need more *help on particular issues.
• So if the student female population is real small, need faculty (male and female) doing the support. ..

Her talk is a paper in the proceedings. Well worth a read.

At PROSIM, I saw many reasonably sophisticated models of parts of the software process. But those folks generally WRITE models, and rarely RUN them.

At the empirical software engineering panel, heard lots of nonsense about "no background theory of software" or "isolated results hard to map into a big picture". So, why not use the PROSIM models as the background theories?

• Monte Carlo simulations of the models;
• Machine learning to find the key controllers;
• Constrain the simulations to the (limited) amount of available data, etc etc.

Many Australians here but no hot news from the land of sunshine.

The Australian government continues to erode research funding; many anecdotal stories of "brain drain" (but no hard figures).

So I'm in the United States for a while.