It wouldn't be uncalled for. The past ten years' worth of patches have pretty much come from Edward Shishkin. I can only imagine the work required to keep such a large codebase maintained out of kernel during that time.

The ideas described in the linked email looks quite original, too. The abstraction level between blocks and files have always been a leaky one. It would been interesting to see them implemented. I guess it would require a lot more RAM to be effective but that might not be an impediment these days.