diff --git a/docs/cerberus.rst b/docs/cerberus.rst
new file mode 100644
index 0000000..bc53789
--- /dev/null
+++ b/docs/cerberus.rst
@@ -0,0 +1,177 @@
+``glom.Match`` compared with ``Cerberus``
+=========================================
+
+Comparing glom Match with the Cerberus library, by
+side-by-side implementation of examples.
+
+
+
+Basic Usage
+-----------
+
+
+.. code-block:: python
+
+
+    >>> document = {'name': 'john doe'}
+
+    >>> schema = {'name': {'type': 'string'}}
+    >>> v = Validator(schema)
+    >>> v.validate(document)
+    True
+
+    >>> spec = {'name': str}
+    >>> m = Match(spec)
+    >>> m.matches(document)
+    True
+
+
+
+Nesting
+-------
+
+
+.. code-block:: python
+
+
+    schema = {
+        'person': {
+            'type': 'dict',
+            'schema': {
+                'name': {'type': 'string', 'required': True, 'minlength': 2},
+                'age': {'type': 'integer', 'required': True, 'min': 0},
+            }
+        }
+    }
+
+    spec = {
+        'person': {
+            'name': And(str, M(T.__len__() > 2))},
+            'age': And(int, M(T > 0)),
+        }
+    }
+
+
+
+Validating User Types
+---------------------
+Cerberus `apparently requires`_ a custom subclass of
+``Validator`` with a ``_validate_type_<typename>``
+method.
+
+
+.. code-block:: python
+
+
+    class UserClass: pass
+
+    class UserValidator:
+        def _validate_type_userclass(self, field, value):
+            if not isinstance(value, UserClass):
+                self._error(field, ERROR_BAD_TYPE.format('UserClass'))
+
+
+.. _apparently requires: https://cerberus-sanhe.readthedocs.io/customize.html#new-types
+
+
+None of this is needed in ``glom`` -- because python types match
+instances of that type, ``UserClass`` can simply be used directly
+in a spec.
+
+
+.. code-block:: python
+
+
+    UserClass  # value is an instance of user class
+    [UserClass]  # value is a list of instances of user class
+    {'data': UserClass}  # value['data'] is an instance of user class
+
+
+Allow Extra Keys
+----------------
+
+
+Cerberus has a special keyword argument ``allow_unknown``
+which is used to let the validator know that extra keys are
+allowed.
+
+``glom.Match`` because it can use types-as-keys, can simply
+express "any key is allowed" with ``{object: object}``.
+
+Moreover, ``glom.Match`` is much more flexible -- any condition
+that can be applied to a value can also be applied to a key.
+
+For example, maybe only string keys are allowed: ``{str: object}``.
+
+Or, maybe anything but ``None`` is allowed as a key: ``{Not(None): object}``.
+
+
+
+.. code-block:: python
+
+
+    >>> v = Validator()
+    >>> v.schema = {}
+    >>> v.allow_unknown = True
+    >>> v.validate({'name': 'john', 'sex': 'M'})
+    True
+
+    >>> Match({object: object})
+
+
+
+Validating Nested Structures
+----------------------------
+
+
+.. code-block:: python
+
+
+    >>> glom.Match({'outer': {str: int}}).validate({'outer': {'a': 1, 'b': 2}})
+
+
+Extension
+---------
+
+Of note here is that glom is very careful about managing global modifications.
+
+
+
+.. code-block:: python
+
+
+    # cerberus custom validator
+    from cerberus import Validator
+
+    class MyValidator(Validator):
+        def _validate_is_odd(self, constraint, field, value):
+            """ Test the oddity of a value.
+
+            The rule's arguments are validated against this schema:
+            {'type': 'boolean'}
+            """
+            if constraint is True and not bool(value & 1):
+                self._error(field, "Must be an odd number")
+
+    # call the validator
+    schema = {'amount': {'is odd': True, 'type': 'integer'}}
+
+
+    # glom custom validator
+    from glom import MatchError
+
+    class IsOdd:
+        def glomit(self, target, scope):
+            if not value & 1:
+                raise MatchError("{0} is not odd", target)
+
+    # call the validator
+    from glom import Match
+
+    spec = Match({'amount': IsOdd()})
+
+    # M and T in glom also allow these kind of expressions
+    spec = M(T['amount'] & 1)
+
+
+