At the present time, we see two big disadvantages of Java: (1) it seems to be 5-30 times slower than Fortran or C, (2) it is immature and (in our experience) embedded in unstable environments.
Scott Morton, Cray's representative to the geophysical world, shared with us his thoughts on a prototyper's philosophy including opinions of Arch Robison (designer of KAI's C++ optimizer). As for performance of Java, Robison expected it to match C by the year 2000. If his company does Java, he said, it will run faster than C++. He forecast that in 1997 there would be more academic conference papers of the my-extension-to-Java kind than of the my-extension-to-C++ kind. For 1998 he predicts that Java will predominate introductory computing languages for computer science majors. Robison was quoted saying that in 1996 local bookstore shelf space for Java exceeds that of Fortran, and in 1997 it will exceed that of C++. By 1997 the tools for Java will be so much better than the C++ tools that the only reason for C, C++, and F90 will be performance and legacy codes. By the year 2000, time to develop applications should be so much better in Java than in C++ that Java shops will have better survival statistics. By 2001 more parallel programs will be written in Java than in any other language.
We are delighted to find Mr. Robison, a powerful figure in the world of compilers, confidently confirming that our dreams are just around the corner, but 2001 is not here yet, and we seek to define now our first steps beyond F77.