Unfinished work: Difference between revisions

From xfs.org
No edit summary
Line 12: Line 12:
== Support for unicode / utf8 filesystems ==
== Support for unicode / utf8 filesystems ==


For normal Unix filesystems the characterset does not matter because filenames are treated as opaque blobs with special treatment only for ".", "..", "/" and "\0".  But for case insensitive filenames the case mapping depends on the locale.  The best way to implement case insensitive filenames is thus to chose one locale and stick to it for the on disk format.  Utf8 has the boadest coverage so it's the best choice if you only want to support one.  Barry Naujok has worked on it and published various patches for this.  The latestet unicode support patch for XFS is here [http://marc.info/?l=linux-xfs&m=120876737022153&w=2]
For normal Unix filesystems the characterset does not matter because filenames are treated as opaque blobs with special treatment only for ".", "..", "/" and "\0".  But for case insensitive filenames the case mapping depends on the locale.  The best way to implement case insensitive filenames is thus to chose one locale and stick to it for the on disk format.  Utf8 has the boadest coverage so it's the best choice if you only want to support one.  Barry Naujok has worked on it and published various patches for this.  The latestet unicode support patch for XFS is here [http://marc.info/?l=linux-xfs&m=120876737022153&w=2] There is also a slightly older patch that can convert from different user charsets to unicode for the on disk format [http://marc.info/?l=linux-xfs&m=120716478703694&w=2]

Revision as of 18:37, 21 December 2008

This is a list of patches that haven't been merged because they haven't been fully completed.

Inline data in inodes

IWAMOTO Toshihiro had a rough patch implementing storing regular file data inside the inode data fork, similar to what we do for small directories or symlinks [1]. Needs more work, and xfs-cmds support.

The xfs_reno tool

The xfs_reno tool has been available for IRIX for a long time. It can be used to make sure all inodes fit into 32 bits which is required for many old applications on 32 bit systems. It also is one of the building blocks required to support filesystem shrinks. Two versions [2] [3] have been posted by Barry Naujok. The current version is enough for it's original usecase, but not yet enough for the shrink functionality. See the required additional steps here [4] for full shrink support.

Support for unicode / utf8 filesystems

For normal Unix filesystems the characterset does not matter because filenames are treated as opaque blobs with special treatment only for ".", "..", "/" and "\0". But for case insensitive filenames the case mapping depends on the locale. The best way to implement case insensitive filenames is thus to chose one locale and stick to it for the on disk format. Utf8 has the boadest coverage so it's the best choice if you only want to support one. Barry Naujok has worked on it and published various patches for this. The latestet unicode support patch for XFS is here [5] There is also a slightly older patch that can convert from different user charsets to unicode for the on disk format [6]