I've resisted joining in this dialogue so far, but here goes...
Jeff Squyres wrote:
> Yes, this is a painful process. And the resulting names won't be as
> simple as they could be. But look at one of the most popular operating
> system in the world -- MS Windows (ouch, it really hurts to say that!).
> Their function, variable, and constant names are extremely long, but
> extremely descriptive. There's no doubt in the user's mind what that
> function/variable/constant is for. I seem to recall that SOM is
An important question here is "who are these users?". I believe that
most users of the MPI-2 I/O functions will be scientific users who are
accustumed to programming in Fortran and C, with some also experienced
in C++. Most don't have any experience doing O/S programming
(including MS Windows), but they *have* opened files. The Fortran
open function is "OPEN", even when using some of the more baroque
high-performance vendor-specific i/o functions. C gives you the
option of "open" and "fopen". I contend that users coming from this
background will find MPI_Open the most intuitive of the possible names
that have been mentioned.
John May followed up:
> Of course, a user who just tries to find the function in the index
> may be looking for a long time ;-)
Agreed. Names of functions should be reasonably obvious to bright
people who are *not* already experts in the library. They ought to
suggest what the function does to the user (MPI_Comm_file_open fails
to do this - it may even lead users to suspect that they hasn't found
the function they were looking for), and be easy to find in an index
(MPI_Open is a moderate win over MPI_File_open in this regard).
-Ian
-- Ian E. Stockdale MRJ Technology Solutions, NASA Ames Research Center ies@nas.nasa.gov (use ies@acm.org for non-work email)