diff --git a/doc/todo/BACKEND b/doc/todo/BACKEND
new file mode 100644
index 00000000000..6184f066006
--- /dev/null
+++ b/doc/todo/BACKEND
@@ -0,0 +1,23 @@
+- documentation of various new functions and the new struct _BackendData
+- add tests for new behaviour and code
+- update existing tests where possible
+- separate `KDB_BOOTSTRAP_STORAGE` from `KDB_DEFAULT_STORAGE`
+  - `KDB_BOOTSTRAP_STORAGE` is used for the hardcoded bootstrap backends at /elektra
+  - `KDB_DEFAULT_STORAGE` is used for the fallback backends at the namespace roots
+  - maybe a complete backend plugin should be defined instead of a resolver+storage
+- reserve /elektra/... in every namespace
+  - keys below /elektra should **always** be stored in the bootstrap backend for the respective namespace
+  - maybe /elektra/mountpoints should use a separate mountpoint so the actual mountpoints are always separate from the rest 
+- implement user:/elektra/mountpoints and dir:/elektra/mountpoints
+- implement cache with new backend logic
+- implement notifications with new backend logic
+- change `spec` plugin, so it can run twice during kdbSet (for adding metadata to new keys and for removing metadata)
+- handle cascading, proc, meta, spec, default keys/mountpoints as described in mountpoints.md
+- implement new mount tooling
+  - The tooling should be able to adapt to different backend plugins.
+    -> Some of the code must be contributed by the plugin.
+  - There should be standard kdb mount CLI syntax similar to what we have now.
+    This syntax should be preferred by backend plugins over a custom solution, if possible.
+    -> There has to be a library with the implementation of this standard approach.
+  - Especially for embedded systems, it should be possible to separate the tooling code from the runtime code.
+    -> Tooling code should live in a separate binary, or at least delegation to a separate binary should be possible.