ISO/IEC JTC1/SC22/WG5-N1931

                          IFIP WG 2.5 Liaison Report
                                Van Snyder

IFIP WG 2.5 convened a workshop on uncertainty quantification in scientific
computing at the Millennium Harvest Hotel in Boulder last August.  Andrew
Dienstfrey, from the NIST laboratory in Boulder was the host.  See
http://www.nist.gov/itl/math/ifip-woco-10.cfm.  Brian Smith, an X3J3
alumnus, gave an interesting presentation on software he developed, that he
calls a "test harness."  The intended use is to test Fortran programs.
Under user control, it inserts instrumentation output statements into a
program.  If desired, one can compare results of instrumentation output, at
each instrumented point in the program, with results from a "gold brick"
run, to verify that nothing has been broken.

This year's meeting and workshop, "Numerical Software: Design, Analysis and
Verification, will be held at the University of Cantabria, in Santander,
Spain, on 4-6 July.  It is hosted by Javier Segura.

To support the task of epistemic uncertaintainty quantification, WG 2.5 is
still keen to have support for interval arithmetic, but realizes that the
computational mathematics community hasn't yet decided exactly what that
means.  It was premature for WG5 and J3 to take up this topic in 1997.
After (if) IEEE P 1788 finishes their work, and when (if) hardware
manufacturers take up the challenge to provide support in hardware, it
might be appropriate for WG5 and PL22.3 to visit the issue again -- but not
before.