Problem Solving Environments: Expectations and Reality

1
Problem Solving Environments Expectations and
Reality

• Richard Fateman
• Computer Science Division
• University of California, Berkeley

2
What do the names mean to us?

• Accelerated, Strategic, Grand Challenge, High
  Performance Environments
• Old wine in new bottles? -- Not really

3
Reality Check

• It is hard to use computers, still
• CS tends to be driven by technology not
  application
• User communities define PSE funding today
• Moores law provides a low-effort quick fix to
  many problems why change approach?
• Yet, we believe
• There are unexploited major opportunities to
  vastly improve productivity.

4
Times Change

• Economic forces Commodity computing replaces
  supercomputing
• Market focus Java? GUI commercial software?
• Technological forces
• standards winner takes all
• ubiquitous networking
• Expertise is distributed, sparse
• Knowledge is everywhere.

5
Whats a PSE?

• Broadly defined to include
• General (Gallopoulos, Houstis , Rice)
• Meta tools - subroutines, interfaces
• Meta2 tool builders (app languages, buses,
  CORBA, html generator...)
• Meta3 system languages (Java, Perl, Tcl, Lisp)
  
• Specific / custom built
• MS Powerpoint, Travellers aid, Purchasing
  advisor, Electronic Notebook, Airframe CAD,
  Biodynamics framework, Wood structures
• Stake in the ground principle
• Easier to sell

6
Technology I ( and -)

• Java is good because although it's usually
  interpreted and is slow it
• Runs really fast on someone else's machine,
• so we hear.
• Java is good because although it is
  object-oriented, it
• Runs really fast on someone else's machine,
• so we hear.

7
Technology II

• Java is good because it is secure,
• and a program written in pure Java is easy to
  debug, has a
• well-defined semantics, and
• Runs really fast on someone else's machine,
• so we hear.
• Java does automatic storage allocation, but
  that's ok because times are different
  now...although garbage collection was funny in
  lisp, GC in Java
• Runs really fast on someone else's machine,
• so we hear.
• Java is really truly machine independent,
  provided all incorrect implementations are erased
  and the correct JVM downloaded.
• Then everyone's Java will compute the same
  thing, so we hear...

8
What about Computer Algebra?

• The right level of discourse for science
• richer than Fortran, C, etc.
• Communicating about Math over networks, stored in
  digital form, etc.
• Extensible library perspective on knowledge
• Framework for programming or meta-programming