Software metrics and software estimation

In order to manage a project we must be able to estimate the time it will take to develop software. This requires breaking a piece of software down into components for which the programmer / developer effort can be estimated. From such estimates we can build up a project plan and manager the delivery of software. But how do you estimate development times for software? It will depend on so many factors, including the experience of the programmer, the complexity of the functions the software delivers and the complexity of the environment. It requires that we can in some way measure a piece of software, a problem which has a whole discipline of software metrics behind it. The problems of software measuring were first raised in Fred Brooke's classic book, ‘The Mythical Man Month'. There are now many methods of measuring software and estimation including COCOMO and Function Point Analysis. How objective can we be? Or does software estimation remain a black art?

Question: Explain the problem of software estimation and software metrics. Why is it so difficult? What is the goal of software estimation? Survey the main approaches to software estimation and identify some common measurable elements. Explain how metrics and estimation develops during the project lifecycle. How would you select estimation methods? What do you think is a common sense approach to software metrics and estimating?