diff --git a/podpac/core/data/datasource.py b/podpac/core/data/datasource.py
index d07d130ef..db0142103 100644
--- a/podpac/core/data/datasource.py
+++ b/podpac/core/data/datasource.py
@@ -152,8 +152,8 @@ class DataSource(Node):
     Custom DataSource Nodes must implement the :meth:`get_data` and :meth:`get_native_coordinates` methods.
     """
 
-    source = tl.Any()
-    native_coordinates = tl.Instance(Coordinates)
+    source = tl.Any().tag(readonly=True)
+    native_coordinates = tl.Instance(Coordinates).tag(readonly=True)
     interpolation = interpolation_trait()
     coordinate_index_type = tl.Enum(["list", "numpy", "xarray", "pandas"], default_value="numpy")
     nan_vals = tl.List(allow_none=True)
@@ -170,8 +170,7 @@ class DataSource(Node):
     # when native_coordinates is not defined, default calls get_native_coordinates
     @tl.default("native_coordinates")
     def _default_native_coordinates(self):
-        self.native_coordinates = self.get_native_coordinates()
-        return self.native_coordinates
+        return self.get_native_coordinates()
 
     # this adds a more helpful error message if user happens to try an inspect _interpolation before evaluate
     @tl.default("_interpolation")
diff --git a/podpac/core/data/test/test_interpolate.py b/podpac/core/data/test/test_interpolate.py
index 1250e63ef..5249fedae 100644
--- a/podpac/core/data/test/test_interpolate.py
+++ b/podpac/core/data/test/test_interpolate.py
@@ -167,7 +167,8 @@ def test_init_interpolators(self):
             Interpolation({"default": {"method": "nearest", "params": {"spatial_tolerance": "tol"}}})
 
         # should not allow undefined params
-        interp = Interpolation({"default": {"method": "nearest", "params": {"myarg": 1}}})
+        with pytest.warns(DeprecationWarning):  # eventually, Traitlets will raise an exception here
+            interp = Interpolation({"default": {"method": "nearest", "params": {"myarg": 1}}})
         with pytest.raises(AttributeError):
             assert interp.config[("default",)]["interpolators"][0].myarg == "tol"
 
diff --git a/podpac/core/data/test/test_types.py b/podpac/core/data/test/test_types.py
index 850791b89..fc3031b48 100644
--- a/podpac/core/data/test/test_types.py
+++ b/podpac/core/data/test/test_types.py
@@ -172,29 +172,10 @@ def test_auth_session(self):
         assert node.auth_session is None
 
     def test_dataset(self):
-        """test dataset attribute and traitlet default """
+        """test dataset trait """
         self.mock_pydap()
 
         node = PyDAP(source=self.source, datakey=self.datakey)
-
-        # override/reset source on dataset opening
-        node._open_dataset(source="newsource")
-        assert node.source == "newsource"
-        assert isinstance(node.dataset, pydap.model.DatasetType)
-
-    def test_source(self):
-        """test source attribute and trailet observer """
-        self.mock_pydap()
-
-        node = PyDAP(source=self.source, datakey=self.datakey, native_coordinates=self.coordinates)
-
-        # observe source
-        node._update_dataset(change={"old": None})
-        assert node.source == self.source
-
-        output = node._update_dataset(change={"new": "newsource", "old": "oldsource"})
-        assert node.source == "newsource"
-        assert node.native_coordinates == self.coordinates
         assert isinstance(node.dataset, pydap.model.DatasetType)
 
     def test_get_data(self):
@@ -298,13 +279,6 @@ def test_dataset(self):
             RasterReader = rasterio.io.DatasetReader  # Rasterio >= v1.0
         assert isinstance(node.dataset, RasterReader)
 
-        # update source when asked
-        with pytest.raises(rasterio.errors.RasterioIOError):
-            node.source = "assets/not-tiff"
-            node._open_dataset()
-
-        assert node.source == "assets/not-tiff"
-
         node.close_dataset()
 
     def test_default_native_coordinates(self):
@@ -349,19 +323,6 @@ def test_get_band_numbers(self):
         assert isinstance(numbers, np.ndarray)
         np.testing.assert_array_equal(numbers, np.arange(3) + 1)
 
-    def test_source(self):
-        """test source attribute and trailets observe"""
-
-        node = Rasterio(source=self.source)
-        assert node.source == self.source
-
-    def test_change_source(self):
-        node = Rasterio(source=self.source)
-        assert node.band_count == 3
-
-        node.source = self.source.replace("RGB.byte.tif", "h5raster.hdf5")
-        assert node.band_count == 1
-
 
 class TestH5PY(object):
     source = os.path.join(os.path.dirname(__file__), "assets/h5raster.hdf5")
diff --git a/podpac/core/data/types.py b/podpac/core/data/types.py
index 0a425eb94..c9cd87b52 100644
--- a/podpac/core/data/types.py
+++ b/podpac/core/data/types.py
@@ -74,7 +74,7 @@ class Array(DataSource):
     `native_coordinates` need to supplied by the user when instantiating this node.
     """
 
-    source = ArrayTrait()
+    source = ArrayTrait().tag(readonly=True)
 
     @tl.validate("source")
     def _validate_source(self, d):
@@ -123,16 +123,17 @@ class PyDAP(DataSource):
         auth_session instead if you have security concerns.
     """
 
-    # required inputs
-    source = tl.Unicode(default_value="")
+    source = tl.Unicode().tag(readonly=True)
+    dataset = tl.Instance("pydap.model.DatasetType").tag(readonly=True)
+
+    # node attrs
     datakey = tl.Unicode().tag(attr=True)
 
-    # optional inputs and later defined traits
-    auth_session = tl.Instance(authentication.Session, allow_none=True)
+    # optional inputs
     auth_class = tl.Type(authentication.Session)
-    username = tl.Unicode(None, allow_none=True)
-    password = tl.Unicode(None, allow_none=True)
-    dataset = tl.Instance("pydap.model.DatasetType")
+    auth_session = tl.Instance(authentication.Session, allow_none=True)
+    username = tl.Unicode(default_value=None, allow_none=True)
+    password = tl.Unicode(default_value=None, allow_none=True)
 
     @tl.default("auth_session")
     def _auth_session_default(self):
@@ -157,7 +158,7 @@ def _auth_session_default(self):
         return session
 
     @tl.default("dataset")
-    def _open_dataset(self, source=None):
+    def _open_dataset(self):
         """Summary
         
         Parameters
@@ -170,23 +171,17 @@ def _open_dataset(self, source=None):
         TYPE
             Description
         """
-        # TODO: is source ever None?
-        # TODO: enforce string source
-        if source is None:
-            source = self.source
-        else:
-            self.source = source
 
         # auth session
         # if self.auth_session:
         try:
-            dataset = pydap.client.open_url(source, session=self.auth_session)
+            dataset = self._open_url()
         except Exception:
             # TODO handle a 403 error
             # TODO: Check Url (probably inefficient...)
             try:
                 self.auth_session.get(self.source + ".dds")
-                dataset = pydap.client.open_url(source, session=self.auth_session)
+                dataset = self._open_url()
             except Exception:
                 # TODO: handle 403 error
                 print ("Warning, dataset could not be opened. Check login credentials.")
@@ -194,22 +189,8 @@ def _open_dataset(self, source=None):
 
         return dataset
 
-    @tl.observe("source")
-    def _update_dataset(self, change=None):
-        if change is None:
-            return
-
-        if change["old"] == None or change["old"] == "":
-            return
-
-        if self.dataset is not None and "new" in change:
-            self.dataset = self._open_dataset(source=change["new"])
-
-        try:
-            if self.native_coordinates is not None:
-                self.native_coordinates = self.get_native_coordinates()
-        except NotImplementedError:
-            pass
+    def _open_url(self):
+        return pydap.client.open_url(self.source, session=self.auth_session)
 
     @common_doc(COMMON_DATA_DOC)
     def get_native_coordinates(self):
@@ -279,14 +260,16 @@ class CSV(DataSource):
         Raw Pandas DataFrame used to read the data
     """
 
-    source = tl.Unicode()
+    source = tl.Unicode().tag(readonly=True)
+    dataset = tl.Instance(pd.DataFrame).tag(readonly=True)
+
+    # node attrs
+    dims = tl.List(default_value=["alt", "lat", "lon", "time"]).tag(attr=True)
     alt_col = tl.Union([tl.Unicode(), tl.Int()]).tag(attr=True)
     lat_col = tl.Union([tl.Unicode(), tl.Int()]).tag(attr=True)
     lon_col = tl.Union([tl.Unicode(), tl.Int()]).tag(attr=True)
     time_col = tl.Union([tl.Unicode(), tl.Int()]).tag(attr=True)
     data_col = tl.Union([tl.Unicode(), tl.Int()]).tag(attr=True)
-    dims = tl.List(default_value=["alt", "lat", "lon", "time"]).tag(attr=True)
-    dataset = tl.Instance(pd.DataFrame)
 
     def _first_init(self, **kwargs):
         # First part of if tests to make sure this is the CSV parent class
@@ -397,9 +380,10 @@ class Rasterio(DataSource):
     * Linux: export CURL_CA_BUNDLE=/etc/ssl/certs/ca-certificates.crt
     """
 
-    source = tl.Union([tl.Unicode(), tl.Instance(BytesIO)])
+    source = tl.Union([tl.Unicode(), tl.Instance(BytesIO)]).tag(readonly=True)
+    dataset = tl.Any().tag(readonly=True)
 
-    dataset = tl.Any(allow_none=True)
+    # node attrs
     band = tl.CInt(1).tag(attr=True)
 
     @tl.default("dataset")
@@ -431,25 +415,6 @@ def close_dataset(self):
         """
         self.dataset.close()
 
-    @tl.observe("source")
-    def _update_dataset(self, change):
-        if hasattr(self, "_band_count"):
-            delattr(self, "_band_count")
-
-        if hasattr(self, "_band_descriptions"):
-            delattr(self, "_band_descriptions")
-
-        if hasattr(self, "_band_keys"):
-            delattr(self, "_band_keys")
-
-        # only update dataset if dataset trait has been defined the first time
-        if trait_is_defined(self, "dataset"):
-            self.dataset = self._open_dataset()
-
-            # update native_coordinates if they have been defined
-            if trait_is_defined(self, "native_coordinates"):
-                self.native_coordinates = self.get_native_coordinates()
-
     @common_doc(COMMON_DATA_DOC)
     def get_native_coordinates(self):
         """{get_native_coordinates}
@@ -672,13 +637,15 @@ class H5PY(DatasetCoordinatedMixin, DataSource):
         Default is 'r'. The mode used to open the HDF5 file. Options are r, r+, w, w- or x, a (see h5py.File).
     """
 
-    source = tl.Unicode()
-    dataset = tl.Any(allow_none=True)
+    source = tl.Unicode().tag(readonly=True)
+    dataset = tl.Any().tag(readonly=True)
+    file_mode = tl.Unicode(default_value="r").tag(readonly=True)
+
+    # node attrs
     datakey = tl.Unicode().tag(attr=True)
-    file_mode = tl.Unicode(default_value="r")
 
     @tl.default("dataset")
-    def _open_dataset(self, source=None):
+    def _open_dataset(self):
         """Opens the data source
         
         Parameters
@@ -691,30 +658,16 @@ def _open_dataset(self, source=None):
         Any
             raster.open(source)
         """
-        # TODO: update this to remove block (see Rasterio)
-        if source is None:
-            source = self.source
-        else:
-            self.source = source
 
         # TODO: dataset should not open by default
         # prefer with as: syntax
-        return h5py.File(source, self.file_mode)
+        return h5py.File(self.source, self.file_mode)
 
     def close_dataset(self):
         """Closes the file for the datasource
         """
         self.dataset.close()
 
-    @tl.observe("source")
-    def _update_dataset(self, change):
-        # TODO: update this to look like Rasterio
-        if self.dataset is not None:
-            self.close_dataset()
-            self.dataset = self._open_dataset(change["new"])
-        if trait_is_defined(self, "native_coordinates"):
-            self.native_coordinates = self.get_native_coordinates()
-
     @common_doc(COMMON_DATA_DOC)
     def get_data(self, coordinates, coordinates_index):
         """{get_data}
@@ -749,10 +702,13 @@ def _find_h5py_keys(obj, keys=[]):
 
 
 class Zarr(DatasetCoordinatedMixin, DataSource):
-    source = tl.Unicode(allow_none=True)
-    dataset = tl.Any()
+    source = tl.Unicode(default_value=None, allow_none=True).tag(readonly=True)
+    dataset = tl.Any().tag(readonly=True)
+
+    # node attrs
     datakey = tl.Unicode().tag(attr=True)
 
+    # optional inputs
     access_key_id = tl.Unicode()
     secret_access_key = tl.Unicode()
     region_name = tl.Unicode()
@@ -852,11 +808,13 @@ class WCS(DataSource):
         The coordinates of the WCS source
     """
 
-    source = tl.Unicode()
+    source = tl.Unicode().tag(readonly=True)
+    wcs_coordinates = tl.Instance(Coordinates).tag(readonly=True)  # default below
+
+    # node attrs
     layer_name = tl.Unicode().tag(attr=True)
     version = tl.Unicode(WCS_DEFAULT_VERSION).tag(attr=True)
     crs = tl.Unicode(WCS_DEFAULT_CRS).tag(attr=True)
-    wcs_coordinates = tl.Instance(Coordinates)  # default below
 
     _get_capabilities_qs = tl.Unicode("SERVICE=WCS&REQUEST=DescribeCoverage&" "VERSION={version}&COVERAGE={layer}")
     _get_data_qs = tl.Unicode(
@@ -1188,7 +1146,9 @@ class ReprojectedSource(DataSource):
         Coordinates where the source node should be evaluated. 
     """
 
-    source = NodeTrait()
+    source = NodeTrait().tag(readonly=True)
+
+    # node attrs
     source_interpolation = interpolation_trait().tag(attr=True)
     reprojected_coordinates = tl.Instance(Coordinates).tag(attr=True)
 
@@ -1270,9 +1230,12 @@ class Dataset(DataSource):
         For example, if the data contains ['lat', 'lon', 'channel'], the second channel can be selected using `extra_dim=dict(channel=1)`
     """
 
+    source = tl.Any(allow_none=True).tag(readonly=True)
+    dataset = tl.Instance(xr.Dataset).tag(readonly=True)
+
+    # node attrs
     extra_dim = tl.Dict({}).tag(attr=True)
     datakey = tl.Unicode().tag(attr=True)
-    dataset = tl.Instance(xr.Dataset)
 
     @tl.default("dataset")
     def _dataset_default(self):
diff --git a/podpac/core/node.py b/podpac/core/node.py
index 637fc0111..6390bc602 100644
--- a/podpac/core/node.py
+++ b/podpac/core/node.py
@@ -139,8 +139,18 @@ def _validate_units(self, d):
 
     def __init__(self, **kwargs):
         """ Do not overwrite me """
+
         tkwargs = self._first_init(**kwargs)
 
+        # make tagged "readonly" and "attr" traits read_only, and set them using set_trait
+        # NOTE: The set_trait is required because this sets the traits read_only at the *class* level;
+        #       on subsequent initializations, they will already be read_only.
+        for name, trait in self.traits().items():
+            if trait.metadata.get("readonly") or trait.metadata.get("attr"):
+                if name in tkwargs:
+                    self.set_trait(name, tkwargs.pop(name))
+                trait.read_only = True
+
         # Call traitlest constructor
         super(Node, self).__init__(**tkwargs)
         self.init()
diff --git a/podpac/core/test/test_node.py b/podpac/core/test/test_node.py
index 0d17c3111..dfcdce9f2 100644
--- a/podpac/core/test/test_node.py
+++ b/podpac/core/test/test_node.py
@@ -314,9 +314,9 @@ def d2(self):
         assert t2.c2() == 2
         assert t2.d2() == 2
 
-        t.a = 2
+        t.set_trait("a", 2)
         assert t.a2() == 4
-        t.b = 2
+        t.set_trait("b", 2)
         assert t.b2() == 4  # This happens because the node definition changed
         t.rem_cache(key="*", coordinates="*")
         assert t.c2() == 2  # This forces the cache to update based on the new node definition
@@ -332,9 +332,9 @@ def d2(self):
         assert t2.c2() == 2
         assert t2.d2() == 2
 
-        t2.a = 2
+        t2.set_trait("a", 2)
         assert t2.get_cache("a2") == 4  # This was cached by t
-        t2.b = 2
+        t2.set_trait("b", 2)
         assert t2.get_cache("c2") == 4  # This was cached by t
         assert t2.get_cache("d2") == 2  # This was cached by t
 
@@ -385,9 +385,9 @@ def d2(self):
         assert t2.c2() == 2
         assert t2.d2() == 2
 
-        t.a = 2
+        t.set_trait("a", 2)
         assert t.a2() == 4
-        t.b = 2
+        t.set_trait("b", 2)
         assert t.b2() == 4  # This happens because the node definition changed
         t.rem_cache(key="*", coordinates="*")
         assert t.c2() == 2  # This forces the cache to update based on the new node definition
diff --git a/podpac/datalib/gfs.py b/podpac/datalib/gfs.py
index f6fa3d00a..4b49c4ef5 100644
--- a/podpac/datalib/gfs.py
+++ b/podpac/datalib/gfs.py
@@ -31,7 +31,6 @@ class GFSSource(Rasterio):
     date = tl.Unicode().tag(attr=True)
     hour = tl.Unicode().tag(attr=True)
     forecast = tl.Unicode().tag(attr=True)
-    dataset = tl.Any()
 
     def init(self):
         self._logger = logging.getLogger(__name__)
diff --git a/podpac/datalib/intake.py b/podpac/datalib/intake.py
index 723caff7a..0be222f4b 100644
--- a/podpac/datalib/intake.py
+++ b/podpac/datalib/intake.py
@@ -57,8 +57,8 @@ class IntakeCatalog(podpac.data.DataSource):
     """
 
     # input parameters
+    source = tl.Unicode().tag(readonly=True)
     uri = tl.Unicode()
-    source = tl.Unicode()
 
     # optional input parameters
     field = tl.Unicode(default_value=None, allow_none=True)
diff --git a/podpac/datalib/smap.py b/podpac/datalib/smap.py
index 20cf30c4e..572c8eab1 100644
--- a/podpac/datalib/smap.py
+++ b/podpac/datalib/smap.py
@@ -445,7 +445,7 @@ class SMAPProperties(SMAPSource):
 
     file_url_re = re.compile(r"SMAP.*_[0-9]{8}T[0-9]{6}_.*\.h5")
 
-    source = tl.Unicode()
+    source = tl.Unicode().tag(readonly=True)
 
     @tl.default("source")
     def _property_source_default(self):
diff --git a/podpac/datalib/terraintiles.py b/podpac/datalib/terraintiles.py
index 9321bcbb2..6fe53b21f 100644
--- a/podpac/datalib/terraintiles.py
+++ b/podpac/datalib/terraintiles.py
@@ -76,14 +76,12 @@ class TerrainTilesSource(Rasterio):
     """
 
     # parameters
-    source = tl.Unicode()
+    source = tl.Unicode().tag(readonly=True)
 
     # attributes
     interpolation = interpolation_trait(
         default_value={"method": "nearest", "interpolators": [RasterioInterpolator, ScipyGrid, ScipyPoint]}
-    )
-
-    dataset = tl.Any()
+    ).tag(readonly=True)
 
     @tl.default("dataset")
     def open_dataset(self):