[devit]

Can anyone explain what's so complex about a "payroll system"?

Presumably you just have a table with people with salary, and every month you wire the money and record information about the wire or otherwise have a way to record when the person has taken cash.

You might want to have more complex rules for determining schedule and amount of salaries, and support for contractors, unpaid leave, taxes, etc. but that doesn't seem particulary complex either.

How can this possibly take more than 100 person-years (i.e. < $30 million) to make even a truly extravagant version of?

Not to mention that presumably such software already exists and could be used instead of writing a new one.

[obviously my explanation would be corruption, but I wonder if I miss something]

[wvanbergen]

Start by reading this: http://wiki.c2.com/?WhyIsPayrollHard

[gerbilly]

Remembering and tracking all the agreements made with workers and unions going back decades and handling the cases where workers change jobs and union designations (or even move from one union to another).

There are thousands of byzantine rules in the contracts and agreements, some mutually contradictory.

The old code likely contained a lot of business logic that represented an institutional memory of how to handle all these corner cases.

Anyone wishing to reimplement a system like that will have to rediscover these corner cases one at a time in production, while thousands of people complain that their paycheque is short.

Then there are also pensions to keep track of etc...

[Chyzwar]

That why you should start small but cover most common cases first. Then gradually work on policies to simplify byzantine rules. Retrain some people and keep internal dev team.

IBM promised a big bang that would be finished before the election. What could possibly go wrong?

I think that there is a positive aspect. Other governments will learn from mistakes when ordering expensive projects like this.