Theoretical Query: POSIX vs NFSv4 ACLs

My attention has been drawn to a Reddit / TrueNAS thread, where someone is having problems with POSIX style ACLs, not knowing about the masks. I don’t know about POSIX masks either.

I took a look at my own TN setup, & ALL my Datasets have NFSv4 ACLs.
I don’t remember specifically setting up NFS ACLs, however I do specify protocol intention with the Dataset. Most of my Datasets are purely SMB. One is both SMB & NFS. ie, my Datasets are each paired with an SMB share, or one is a multi-protocol Dataset: SMB & NFS shares paired.

If I ask my browser AI about TN POSIX vs NFS ACLs, I get the info that POSIX is TN default, & I would need to change the ZFS ACL default to NFS on the Dataset if I want that.

It seems that this AI data is old. Seems that TN now favours NFS ACLs for Datasets intended for either SMB or NFS shares.

What are others running here? POSIX or NFS ACLs?
I’m not gonna change my config. It works. But curious what other TN admins are doing.

I use POSIX. AFAIK, NFS are only required for advanced setups.

Is this pool originally from a TrueNAS Core or FreeNAS system? If so, ZFS on FreeBSD only supports acltype=nfsv4.

It’s not really advanced setups as it is what you need. NFSv4 when set properly makes it really easy to ignore that permissions are ever a thing (just declare it so that groups have MODIFY permissions inheriting on everything and don’t bother with owner@, group@, everyone@ entries). Leave aclmode to “passthrough” so that apps that do internal chmod are left with the toy permissions steering-wheel but can never actually change permissions. I have several home server SMB shares I set permissions on maybe 7-8 years ago and haven’t touched since.

I only use POSIX. My laptop is a Macbook, so I don’t have to fight the Windows SMB stuff.