Logged In: YES 
user_id=658986

OK, I know it's not Levenshtein because I've read the
source, and I agree I should really read the docs first and
the source after, it's actually written in the docs.

However, the docs say

`This does not yield minimal edit sequences, but does tend
to yield matches that ``look right'' to people.'

This is not true -- see my last posting.  Or perhaps you
really think matching `Observation' looks better to people
than matching the smaller parts from the rest of the string
(which I strongly doubt).  Then please add a note the
matches it finds can be arbitrarily bad, at least (e.g., the
ratio between optimal and difflib's best match can be as big
as (N-1)^2/4N, where N is sequence length (of both sequecnes)).

Thank you.