[OpenAFS] Feature request, sort of

[Kim Kimball]

Steve Simmons wrote:
>
> On Jul 19, 2007, at 7:10 PM, Kim Kimball wrote:
>
>> Why not restore the volume (restores to RW), replicate it (same 
>> server and partition) and then remove the RW?
>>
>> Mount the resulting readonly explicitly -- i.e. be sure to include 
>> the .readonly suffix in the fs mkm
>>
>> Works for me.
>
> Works perfectly. Also gives the same error reports when you do a 'vos 
> examine' on it. Which is  how this feature request got started.
>
And continued with a discussion of 2x space, users writing into the 
restored volume, etc.

Seems like a lot of work to avoid typing ".readonly" and I don't believe 
a fourth volume type buys us much.

If you want an invariant volume replicate it and delete the RW.  Want to 
write to it restore the RW (dump an RO, convert, whatever), write to it, 
release it, and delete the RW.

The difference between volume.readonly and volume.invariant appears to 
be that 1) there wouldn't be an error from "vos exa" and 2) it wouldn't 
be what, restoreable to RW?  Can't be dumped?  or 3) Certain values on 
ACLs can be disabled -- such as adwi or whatever -- all of which are 
irrelevant on an RO volume anyway.

Not sure I see much benefit.

Kim

> Since then there's been further discussion here that leads me to think 
> a production read-only volume is a good idea. Think about forensics, 
> locking out users in cases of potential but unproven abuse, avoiding 
> per-file callbacks, etc, etc...
>
> Steve
>
>