I thought of one more enhancement that would make diff a killer app: Incorporate the functionality of GNU 'numdiff' into diff.
Numdiff requires the files to be "putatively" similar, meaning that all lines are similar. Numdiff does not have the capability to find similar lines like diff, but if diff could provide the similar lines to numdiff, then its options and algorithm could be applied. At the moment I am using a Perl driver (work in progress) to provide diff output on two files, then use that output to provide line pairs to numdiff for comparison. That process would be obsolete if numdiff were incorporated into diff. Thanks for listening. -Tom
