diff --git a/changes.md b/changes.md
--- a/changes.md
+++ b/changes.md
@@ -1,3 +1,6 @@
+# 0.0.0.2
+    - Improved documentation
+
 # 0.0.0.1
     - Improved documentation
 
diff --git a/filepath-crypto.cabal b/filepath-crypto.cabal
--- a/filepath-crypto.cabal
+++ b/filepath-crypto.cabal
@@ -1,5 +1,5 @@
 name:                filepath-crypto
-version:             0.0.0.1
+version:             0.0.0.2
 synopsis:            Reversable and secure encoding of object ids as filepaths
 license:             BSD3
 license-file:        LICENSE
diff --git a/src/System/FilePath/Cryptographic.hs b/src/System/FilePath/Cryptographic.hs
--- a/src/System/FilePath/Cryptographic.hs
+++ b/src/System/FilePath/Cryptographic.hs
@@ -9,8 +9,11 @@
 type of kind 'Symbol').
 The serialized payload is encrypted using a symmetric cipher in CBC mode using
 the hashed namespace as an initialization vector (IV).
-The ciphertext is then <https://hackage.haskell.org/package/sandi/docs/Codec-Binary-Base32.html base32>-encoded and padding stripped.
 
+The ciphertext is then
+<https://hackage.haskell.org/package/sandi/docs/Codec-Binary-Base32.html base32>-encoded
+and padding stripped.
+
 Rather than being indicated by the amount of padding, the length of the
 serialized plaintext is instead carried at the type level within
 'CryptoFileName' (analogously to the namespace).
@@ -20,9 +23,9 @@
 Since the serialized payload is padded to the length of the next cipher block we
 can detect namespace mismatches by checking that all bytes expected to have been
 inserted during padding are nil.
-The probability of detecting a namespace mismatch is thus
-\[1 - 2^{b \cdot \left \lceil \frac{l}{b} \right \rceil-l}])
-where \(l\) is the length of the serialized payload.
+The probability of detecting a namespace mismatch is thus \(1 - 2^{b \left \lceil \frac{l}{b} \right \rceil - l}\)
+where \(l\) is the length of the serialized payload and \(b\) the length of a
+ciphertext block (both in bits).
 -}
 module System.FilePath.Cryptographic
   ( CryptoID(..)