Splitting diff into library and command-line tool

[Denver Gingerich]

I am interested in spending some time refactoring diff so that it has
two parts: a difference library that performs the actual difference
algorithm, allowing one to abstract out the actual algorithm being
used, and a command-line tool that uses the difference library (this
would be the "diff" command people are used to using).  After the
change, diff would appear to the user to be the same, but it would
allow developers to use the difference library without having to pull
apart the diff code or post-process the diff output.  The library
would be useful for a variety of projects, including version control
systems and other types of diff tools (such as GNU wdiff).

I'm wondering if anyone on the list can provide pointers on the best
way to do this.  Suggestions such as what the API for the difference
library should look like (ie. which functions it would provide) would
be appreciated.  Also, if someone knows of a better algorithm than
Myers' (which I understand diff uses), then I could look into
implementing that as an additional library (so the user can choose
between Myers' and another algorithm).

Denver