Some of ZFS ACLs doesn't work as expected

[Edward Tomasz Napierała]

Wiadomość napisana przez Pavel Bychykhin w dniu 19 sie 2012, o godz. 19:56:
> 19.08.2012 19:40, Edward Tomasz Napierała пишет:
>> Wiadomość napisana przez Pavel Bychykhin w dniu 18 sie 2012, o godz. 19:48:
>>> Dear community!
>>> 
>>> After my experiments with ZFS, I concluded, that permissions "delete_child" and "delete" are ignored.
>>> For the create/update/delete operation a list of "rwxp" (read_data/write_data/execute/append_data) is fully sufficient.
>> 
>> They are not ignored, but yes, write access on a directory is enough to delete a file.
>> 
>>> No need to specify the "delete_child" and "delete" permissions at all, or I don't understand something?
>> 
>> Unless you need them - no, you don't.  That's why these bits are not set in a default
>> case (so called 'trivial ACL', i.e. no ACL set on a file).
>> 
> 
> Could you please provide an example of at least one practical situation, where the "delete_child" and "delete" permissions would be useful?

You could allow for file creation, but deny file removal.  Still, as someone
already mentioned, main reason for these to exist is compatibility with Windows
and NFSv4 spec.  It's just that they are not _completely_ ignored, like SYNCHRONIZE
or READ_XATTR/WRITE_XATTR are.

-- 
If you cut off my head, what would I say?  Me and my head, or me and my body?