diff --git a/include/ada/scheme-inl.h b/include/ada/scheme-inl.h
index 465f33761..aea9c2f57 100644
--- a/include/ada/scheme-inl.h
+++ b/include/ada/scheme-inl.h
@@ -24,14 +24,17 @@ constexpr uint16_t special_ports[] = {80, 0, 443, 80, 21, 443, 0, 0};
 
 /****
  * In is_special, get_scheme_type, and get_special_port, we
- * use a standard hashing technique to find the index of the scheme in the
- * is_special_list. The hashing technique is based on the size of the scheme
- * and the first character of the scheme. It ensures that we do at most one
- * string comparison per call. If the protocol is predictible (e.g., it is
- * always "http"), we can get a better average performance by using a a
- * simpler approach where we loop and compare scheme with all possible
- * protocol. In this instance, we choose a potentially slightly lower
- * best-case performance for a better worst-case performance.
+ * use a standard hashing technique to find the index of the scheme in
+ * the is_special_list. The hashing technique is based on the size of
+ * the scheme and the first character of the scheme. It ensures that we
+ * do at most one string comparison per call. If the protocol is
+ * predictible (e.g., it is always "http"), we can get a better average
+ * performance by using a simpler approach where we loop and compare
+ * scheme with all possible protocols starting with the most likely
+ * protocol. Doing multiple comparisons may have a poor worst case
+ * performance, however. In this instance, we choose a potentially slightly
+ * lower best-case performance for a better worst-case performance.
+ * We can revisit this choice at any time.
  *
  * Reference:
  * Schmidt, Douglas C. "Gperf: A perfect hash function generator." More C++