New ZFS "passthrough" ACL inheritance rules [PSARC/2008/231 FastTrack timeout 04/08/2008]

[Mark Shellenbaum]

Darren J Moffat wrote:
> Mark Shellenbaum wrote:
>> Darren J Moffat wrote:
>>> Since this proposed behaviour is the default for ACLs on UFS why 
>>> isn't it the default for ZFS too ?
>>
>> I'm more than willing to make this the default behavior for ZFS, but 
>> it will affect POSIX compliance.  If thats alright with everyone then 
>> I can change it to be the default.
> 
> Would this be the one and only setting that means that a ZFS dataset 
> wouldn't be in a POSIX compliant configuration by default ?  If it is 
> then I would say it shouldn't be changed.  However if there are others 
> then IMO the default ACL behaviour should be the one that matches NFSv4 
> and what people expect of ACLs regardless of what POSIX thinks.
> 

Lets leave the default as it is, and have the new inheritance behavior 
only take place under "passthrough".  We can always change this later 
with another fast track.

> 
>>>
>>> Also shouldn't "secure" be "posix" because "secure" is subjective and 
>>> relative.
>>>
>>
>> secure is what it was called in the original ZFS ARC case.  I can 
>> change it to "posix" if you want.  My only concern would be if users 
>> have become accustomed to its present value.
> 
> What about an alias ?
> 

How about I rename "secure" to "restricted" and have an alias for "secure".

   -Mark