> I understand the proprietary nature of IUCV and therefore would expect IB > M > to write the OCO device drivers
IUCV isn't hard -- that's been done. It's the dataflow verbs and responses that make up the conversation between a SFS/BFS client and the SFS/BFS server that are the undocumented bit. If those were ever documented, I don't think it'd be too long before we had a much better NFS server for CMS data...8-). > for *BLOCKIO *BLOCKIO isn't too bad, actually, as you get to invent your own strategy for requesting and supplying blocks to and from CP. If someone didn't have a requirement that Linux run without VM, *BLOCKIO would have been a much easier method to implement for Linux I/O than the current model.