On Wed, Oct 31, 2018 at 7:39 PM, David Howells <dhowells@xxxxxxxxxx> wrote: >> My objections fundamentally is that I can find real problems when I look >> at the code. I think the big risk with such a change is not that there are bugs, but that we get the API wrong, and have to keep supporting a broken API forever. Were any of your objections of that type? There's the argument about sharing of super blocks doing weird things, but we can't get rid of that one due to having to support the old mount(2) API forever anyway. That's not a new burden, and the new API allows fixing this incrementally by adding better models later. Thanks, Miklos
