[Info-vax] New filesystem mentioned

[Michael Moroney]

Simon Clubley <clubley at remove_me.eisner.decus.org-Earth.UFP> writes:

>On 2019-05-14, Michael Moroney <moroney at world.std.spaamtrap.com> wrote:
[]

>But isn't this a problem in the filesystem drivers instead of in the
>filesystem itself ?

>I understand what you say above, but I am not seeing anything which is
>ODS-2/ODS-5 specific.

It is true that there isn't much on the file system itself that must be cluster
aware. Some data related to the need for a rebuild including number of times
it is mounted in a cluster and maybe a few other fields changed to counts rather
than true/false. For ODS-2/5 just about all that is needed is in the SCB block
and the homeblock as well as certain allocation cache bitmaps.

But the filesystem drivers (RMS/XQP on VMS) is where all the fun is. It wouldn't
be possible to take a 'drop in' Linux driver and put it on VMS and expect it to
be cluster-aware even if VMS had the equivalent interfaces to Linux drop in drivers.
Which it doesn't.

>This seems more about instances of the filesystem drivers running on
>different nodes using the DLM to synchronise caching between themselves
>as opposed to anything that's actually stored in the filesystem on-disk
>structures themselves.

>I don't see anything here that would stop a FAT32 cluster-aware filesystem
>driver (for example) from also using the DLM to coordinate the caching.

I actually poked around with seeing what would be necessary to implement a FAT ACP
as a method for learning the very esoteric (even for VMS) concept of user-written
ACPs. Never got anywhere with it.