This message is from the T13 list server.
I think we are in agreement.
Me too.
I would like to discuss the direction at the
CAP meeting. From my perspective, it is not a big deal to encode the
direction as a part of the opcode. However, I am a bit concerned because we
could end up burning 6 opcodes for this capability (12/16 byte CDB
IN/Out/Non data transfer).
To fit the SCSI tradition of deciding direction by op, we could burn just 2 ops, and leave the ATA-specific distinctions to appear inside the CDB. For example, we could say that an expected length of 0 means non data transfer protocol, no matter if it appears in combination with the SCSI ATA In op or the Out op.
If we only sometimes need more than 12 bytes of CDB, then we sometimes do, and if we want the compatibility of also defining a 12 byte rather than 16 byte CDB, then the cost rises to 4 ops rather than 2, yes. But I've missed the argument for needing to burn 6 ops.
Perhaps we should ask the t10 at t10.org folk to comment on the continuing significance of the decide direction by op tradition.
Pat LaVarre
