Measuring software size by distinct lines

The relationship between DLC (distinct line count) and NCSL (noncomment source lines) is studied on a number of programs, and it is found that, as a simple rule of thumb, the NCSL count can be estimated by twice the DLC. A more accurate model is derived by predicting NCSL from DLC and the number of lines that occur exactly once. It is also shown that, for unrelated programs, the proportion of common lines is very small; hence, DLC is approximately additive. It is concluded that, overall, the DLC is a very attractive measure of size that has two basic advantages over NCSL: it is an intuitively more appealing measure of effort than NCSL, and the problems of measuring size of subsequent releases disappear when using DLC