Re: More IO issues
Jean-Pierre Prost (jpprost@watson.ibm.com)
Thu, 17 Apr 1997 13:47:01 -0400
So the real question is -- what is the *intent* of the MPI io
subcommittee? John May has mentioned that he will be making an IO
implementation that will be independant of the MPI-2 below it. This
seems to sound like the IO group wants it to be layered...?
*****************************
There is a noticeable difference between "layerable" and "layered".
"layerable" makes "layered" possible, not required. I do not think the I/O
subcommitee
wants to require the I/O chapter to be layered. If it was so, the I/O
chapter would not
need to be a chapter. It would just be a library on top of MPI(-2).
John May and others may implement the I/O chapter as a layer on top of the
rest of
MPI-2.
This does not mean that every implementation will do so.
And it is clear to me that a *full* MPI-2 implementation will have to
implement the I/O routines,
may these routines be layered or not.
Now, issues about which libraries, how many of them, which include files,
and how many of them,
can be discussed. I am not sure these should necessarily be implementation
independent though.
Jean-Pierre
****************************