diff --git a/README.md b/README.md
index 599d85c..6aae3c1 100644
--- a/README.md
+++ b/README.md
@@ -2,6 +2,12 @@ __RLTK__
 
 An attempt to manually port some of nltk to rust.
 
+from https://www.nltk.org/api/nltk.lm.html:
+
+_So as to avoid re-creating the text in memory, both train and vocab are lazy iterators. They are evaluated on demand at training time._
+
+rltk has the same philosophy: everything is done using iterators (on iterators) on string slices.
+
 Currently in it's infancy (but growing): 
 * rltk::lm::preprocessing::pad_both_ends(\["a","b","c"], 2) -> "\<s>", "a", "b", "c", "\</s>"]
 * rltk::util::pad_sequence == same as above with customisation
diff --git a/src/metrics/mod.rs b/src/metrics/mod.rs
index fabd061..39954c1 100644
--- a/src/metrics/mod.rs
+++ b/src/metrics/mod.rs
@@ -5,6 +5,8 @@ pub mod distance;
 /// For example, transforming “rain” to “shine” requires three steps, consisting of two substitutions and one insertion:
 /// “rain” -> “sain” -> “shin” -> “shine”.
 /// These operations could have been done in other orders, but at least three steps are needed.
+///
+/// substitution cost is (for now at least) hardcoded as 2
 pub fn edit_distance(s1: &str, s2: &str) -> usize {
     distance::get_edit_distance_table(s1, s2)[s1.len()][s2.len()].value
 }