IBM lingo referring to something on mainframes we would normally refer to as
file.
A
file, by the same convention, would refer to a logical name for a dataset.
An example from TSO:
The command
ALLOC DATASET(SYS1.UADS) FILE(FT07F001)
would assign the dataset SYS1.UADS (or User Attribute Data Set - the file with the TSO user attributes - similar to Unix /etc/passwd) to a logical file FT07F001.
You could then run a FORTRAN program (or COBOL, or PL/I) accessing the file. For FORTRAN, this would have looked like this :
READ(7,100)I,J,FOO,BAR
FORTRAN on MVS did not know about dynamic file assignment - you could not just use an OPEN statement. Thus, the naming convention : FT, followed by the file number, followed by F, followed by an index - starting from 1. (After reading to the end of FT07F001, the program would try to read FT07F002, FT07F003 etc. and only after exhausting all of these, it would finally get an end of file condition.)
Would the above example work ?
No. Of course not. The file system on IBM mainframes is rather different from that on Unix or Windows. A file is not just an unstructured collection of bytes. FORTRAN could only process a rather small subset of the possible combination of file attributes (fixed length records only, ONE of the limitations).
So - even if you knew how to program, you could not access all of the files on your system - unless, of course, you were an Assembler wizard.