(j3.2006) an alternative to Aleks' asynchronous proposal

[Van Snyder]

I have two approaches, outlined in 08-204 and 08-205, that use existing 
ASYNCHRONOUS exactly, without any changes in its meaning, and without 
any additional baggage such as procedure attributes or new meanings of 
existing attributes or statements.

The approach in 08-204 is simplest for programmers (at least for new 
code) and the standard, but more work for processors.

The approach in 08-205 is more work for the standard, but probably not 
much extra work  for processors that already have defined I/O.  It's 
essentially the same as defined I/O, but tied to a unit instead of a 
type.  It's also more work for programmers than 08-204, but I think less 
work than using MPI directly.  I proposed this in principle in 97-114.  
It's probably too much to do now, and I prefer 08-204 anyway.

Van