mysql on ZFS

I don´t know what fixes he mean, but i assume he think about a class of patches introduced to give the user a few more dials to tune and to optimze default behaviour.

This bugfixes are not really *bug*fixes. They change the default behaviour of ZFS in certain cases to run ZFS in a more optimal way (you can enable most of the behaviour changes with tricks before). By the way, we call those tickes change requests because not all patches are bugfixes. ZFS as a filesystem is stable and customers already use ZFS in production at critical systems.

It´s the same with UFS at Solaris. Directio for example wasn´t there from the start. It´s not a technology problem, it´s about the mindset of adopting new technology. There are two mindsets: Starting to adopt when a technology is good for a certain task, or wait and adopt when all issues in even remote cases are solved (independently of the actual occurance at a customer).

The first class of adopters see the inherent advantages of ZFS (ease of use, checksums, guaranteed data vadility) and want these features. So they start to use them, whereever possible, start to learn, and broaden the adoption later.

The second clas od adopters will habe a more steep learning curve, when they make the "we use it when it´s perfect". Not a wise choice from my point of view.