Roland Tanglao takes exception to an AO post. The post...

How efficient is it to pay a software engineer in the Valley a loaded salary of $170,000, the average salary reported in the fourth quarter of 2001, when Asian engineers provide a much better value? We've all read the cost differentials between US and Indian, Vietnamese and Chinese workers. And one of the main reasons this work went overseas is because clients knew they were being gouged by US engineers and consultants. After all, programming is, essentially, production work. And is labor not the most expensive variable component of a software product?

Roland...

Good software is not production work. If this guy had ever actually developed good software, he'd know.

There's a huge fork in software development. If you can define scope clearly and it seems like a straightforward thing to build, then you shop it to a code farm.

On the other hand, if the scope is fuzzy, elastic and frequently changing, and there are elements of novelty (never done that before), sub-cultural awareness (how do our surgical nurses model their work process), you may need local auteurs.

When the business schools started to teach MIS 30 years' ago, you could tell that whole categories of software work would become routine. Both the problem set and the tool set had no barriers to entry. We have a million programmers in North America because of lighter weight problems and easier to use tools. That same lack of barriers makes it easy for India, China, and the rest of the world to enter our labor market.

The things that don't fit?

Thorny problems. Intractable ones that take deep scientific education and grad school maths. Edison problems that require years of tinkering to get the chemistry just right. Collaborative ones that involve close knit teams of world class experts.

Proximity. Where the development team must work intimately with the customer, eating and breathing with them.

And those barriers won't last. When you throw mass quantities of smart people at an education system, like they are doing in India and China, the bell curve says some will become world class computer scientists. People who invent things, who break through conventional thinking, who upset the apple cart. And they will compete with the industrial world's best and brightest.

So you have a few choices.

• Get your PhD in computer science.
• Become a product manager or product requirements manager, because outsourcing demands coordination.
• Become a world class specialist in a technology or an application, one of the top 5 in a very narrow field.
• Or get out.