diff --git a/TODO.libntfs b/TODO.libntfs
index e69de29b..908218d1 100644
--- a/TODO.libntfs
+++ b/TODO.libntfs
@@ -0,0 +1,63 @@
+*****************
+* HIGH priority *
+*****************
+
+- add read of compressed attributes
+
+*******************
+* MEDIUM priority *
+*******************
+
+- create API reference book template (cf. linux kernel)
+- enable automatic creation of API reference
+
+- add write of compressed attributes
+
+- write attribute resize function (see mkntfs) and implement ntfs_truncate*
+
+- write ntfs_{allocate,free}_cluster(s) (see ntfs 2.4 driver in CVS)
+
+- extend ntfs_attr_pwrite to cope with extending the attribute size and with
+  instantiating holes
+
+- write ntfs_{allocate,free}_mft_record (see ntfs 2.4 driver in CVS)
+
+- write ntfs_{rm,create,add}_attr()
+
+- write ntfs_{rm_from,add_to,create,rm}_index()
+
+- write ntfs_file_name_{add,rm}_from_mft_record() or _from_ntfs_inode(?)
+
+- write ntfs_unlink_file()
+
+- write ntfs_create_file()
+
+- write API for conventional high level file access.
+
+- implement loads of utilities a-la ntcp, ntrm, ntcreat, ntdir, etc...
+
+- implement a ntfs shell where can use the above much faster with caching,
+  probably extending the library in the process
+
+- extend attrib API with ntfs_rl_pread()
+
+****************
+* LOW priority *
+****************
+
+- Do we attach attributes (ntfs_attr) to the corresponding ntfs_inode? Now we
+  just atach the inode to the attribute and expect the user to not shoot
+  themselves in the foot.
+
+- add read/set of various file attributes/flags to library
+
+- add ACL read/write support to library
+
+- add MS BackupAPI to library
+
+- add volume resizing support to library
+
+- add defrag API to library
+
+- write utilities for all of the above
+