doc: update design proposal for consuming

sdk/ml/azure-ai-ml/tests/internal/e2etests/test_component.py

@@ -99,3 +99,37 @@ def test_component_load(
 
             # TODO: check if loaded environment is expected to be an ordered dict
             assert pydash.omit(loaded_dict, *omit_fields) == pydash.omit(expected_dict, *omit_fields)
+
+    def test_component_with_default_label(
+        self,
+        client: MLClient,
+        randstr: Callable[[str], str],
+        yaml_path: str,
+    ) -> None:
+        component_name = randstr("component_name")
+
+        create_component(client, component_name, path=yaml_path)
+
+        default_component = client.components.get(component_name)
+
+        # plan A: the version will be set to azureml_default
+        # pros:
+        # 1. easy to implement
+        # 2. unified solution for all component label (azureml_labela, e.g.)
+        # cons:
+        # 1. can't update component based on loaded component
+        # 2. the hidden keyword is exposed to user so not so safe
+        assert default_component.version == "azureml_default"
+
+        # plan B: use a protected field to indicate the label
+        # pros:
+        # 1. can be used to update component
+        # 2. hide keyword from user
+        # 3. unified solution for all component label
+        # pros:
+        # 1. introduce a new "protected" field like "_is_anonymous"
+        assert default_component._label == "default"
+        # or add a special logic in job_operations so all behaviour is hidden from public interface?
+        assert default_component.id == f"{component_name}:azureml_default"
+        assert default_component.version == "1"
+