xref: /src/sys/fs/hfs/TODO

[dillo]

- check _lookup locking (deleted PDIRUNLOCK)
- unicode decomposition
- make type/creator accessible
- mount options: default uid/gid/mode
- return correct error for device not ready (CD) (not Operation not Permitted)
- use lookup cache
- use pool for struct hfspnode
- reduce use of malloc and memcpy
- slim down libhfsp (e.g. replace cbargs with void *)
- union/nfs support
- KNF
- update copyright of apmlabel

Last revision: 2005-08-31


There are HEAPS of limitations and deficiencies in this implementation. In fact,
this code doesn't do a lot more than the very bare minimum needed to even be
called an HFS+ filesystem. Some of the most notable features that need to be
written or improved are, in approximate order of importance:

- uid/gid remapping for non-superusers (possibly)
- support for special nodes
- MASSIVE performance and memory optimization, using lots of caching; this fs is
  about as unoptimized as you can get
- better error handling
- usage of UBC and memory pools
- handling of bad blocks
- HFS wrapper support
- NFS support
- union mount support
- check for broken functionality or security risks due to insufficient range
  checking
- regression tests
- allow volume to be used as root filesystem
- reformat the code to use 8-space tabs :-\


This is, of course, in addition to the missing major features:

- write support
- journal support
- fsck_hfsp, newfs_hfsp
- vanilla HFS support
- man pages and lots of other documentation