"What's the problem?" Stephen J. Mellor, chief scientist at Accelerated Technologies, asked in his Monday afternoon tutorial on model driven architecture (MDA) at SD Best Practices in Boston. "The problem is a perennial one -- software is expensive and productivity is low."

Why?

According to Mellor, who's also a coauthor of MDA Distilled (Addison-Wesley, 2004) and chair of an OMG Analysis and Design Task Force Working Group, it's partially because code exists at too low a level of abstraction. Over the past 30 years, Mellor said, studies have shown that the number of lines of code a person can write during one day is constant -- no matter the language. "So a line of Java and C is obviously more efficient than a line of assembly code," he continued.

But the problems don't stop at the code level; Mellor also believes that reuse occurs at too low a granularity, that there's too much glue code being written, and, worst of all, that the mapping information is lost. "You bind things together based on rules in your head," Mellor explained. "Once you've produced the artifact, how you created it is lost." This loss of information becomes critical as systems grow because "we have to apply that expertise over and over again. Obviously, that does not scale."

[from the Software Development Newsletter, Tamara Carter ]