NicC wrote:Actually, I was told by IBM that a detaset is NOT the same as a file - a file is pure data whereas a dataset is data + control information eg RDWs for VB files, inter-block gaps, start of block markers etc.
I've never seen that definition in print, but ... I think issues like inter-block gaps and start of block markers are present in other systems, but they are hidden by by the system, where us mainframe types are exposed to them. I regard a "file" as nothing more than a stream of bits, where a dataset has something more, though a RECFM F data set is represented as a stream of bytes, but divided into fixed length groups called "records."
This discussion brought a nasty problem with 2314s to my memory. They had something called an "address marker" in the gap between the end of a record and the count area for the next record. The hardware detected this "address marker." If it wasn't there - all too common - you got an I/O error and that track was pretty much done for, though you could usually format write over the track.