Date: Thu, 17 Jun 1999 13:37:03 -0700 (PDT)
From: Linda Stanberry <linda@XXXXXXXXXXX>
To: mpi-core@XXXXXXXXXXX
Subject: Request for interpretation
Cc: bcarnes@XXXXXXXX,hedges1@XXXXXXXX,koniges1@XXXXXXXX,rkyates@XXXXXXXX,Sender: owner-mpi-core@mcs.anl.gov
Precedence: bulk
X-UIDL: dc798902191079a947ed1f06fc1ac02c

I have a question about the interpretation of the wording in Section 4.10, 
Info objects.  The first paragraph describes these objects and indicates
that they are used by many of the routines in MPI-2.  So how do we 
interpret the first sentence of the second paragraph:

	If a function does not recognize a key, it will ignore it, unless
	otherwise specified.

Does this mean that MPI_INFO_SET may ignore (i.e., refuse to add) a key
if the key given is not recognized by any of the MPI-2 functions in this
implementation?

Or does it mean that an MPI-2 function that takes an Info object as an
argument (e.g, MPI_FILE_OPEN or MPI_COMM_SPAWN) will ignore any key in
the object that is not recognized for this function in this implementation?

I would prefer the latter interpretation so that the decision to recognize
or ignore a key in an Info object is delayed until a function uses an Info
object.  With this interpretation, layerability and extensibility of MPI-2
are permitted.

Without this, I am curtailed from layering my own MPI-IO (for example)
over another MPI-2 implementation or from developing new MPI functionality
that uses Info objects with keys that are not recognized by an existing
implementation.  Of course, if I want to do this, I _am_ free to use my own
implementation of Info objects as well, but at the risk of introducing
unintentional side effects to, say, MPI_WIN_CREATE or MPI_COMM_SPAWN in
the host implementation.

Are there other interpretations?  Opinions?

Linda Stanberry
lstanberry@XXXXXXXX