diff algorithm: matching similar lines

Am I right that the old and current darcs-diff algorithm only match 
exactly equal lines? I assume that results are more intuitive if also 
non-matching lines are taken into account depending on their degree of 
similarity (only differences in spacing, single character inserts or 
deletions). Or is this too computationally expensive?

I guess that a spacing-aware diff algorithm might also be a step towards 
an indentation patch type.