The qtree quota, as I see things, has no benefit to acrue from having soft limits - running out of quota in a qtree looks like you've just run right out of disk to a client and folk have learnt to handle that themselves.
You read my mind.
When I first designed the quota tree stuff, I really thought of it as a new form of quota. (Hence the name.)
Over time, we realized (by which I mean our customer kindly informed us) that should behave much more like partitions. So we changed the error when a quota tree fills up from EDQUOT to ENOSPC. And we also changed it so that -- unlike regular quotas -- root can't exceed the quota.
I now wish that we'd called it "virtual volumes," or something else that didn't emphasize the relationship with quotas.
Now I expect that it wouldn't be much harder to apply soft quotas to qtrees than to user and group quotas -- although it might be, since the Sun quota tools have no such concept and also because we've already broken the symetry with user and group as a result of the changes described above.
But I'm a strong believer of "When in doubt, leave it out", so I'm curious whether people really think they would use soft quotas on quota trees, or whether they just like it for the symmetry. (Hey -- I'm a fan of symmetry too.)
If soft quotas on virtual volumes are useful, then should there also be soft quotas on partitions as a whole? (I said I was a fan of symmetry.)
Dave