I have found many interesting things to think about in the
posts on this exploder however I have seen no explicit
reference to what I consider to be the biggest problem in
astronomical software. This is what I shall call the encoding
of knowledge or expertise. I come at this problem from three
related viewpoints: I am the author of a widely-used program,
I use many different tasks in my own research, and I help out
other researchers using a variety of tasks and packages. This
has led me to two related observations about general users :
1) they want to be able to accomplish their research with
the minimum of interaction with the software.
2) they (we) do some really stupid things sometimes.
So, my plea is the following. By all means try to improve the
interoperability of software and all those other good things.
Just don't spend so much of the available resources on it that
there is nothing left to put together high level scripts/tasks
for specific astronomical uses. These scripts/tasks need to
operate with the minimum of user input and they need to have
enough encoded expertise that they catch cases when the user
is probably being stupid. It will do the average user no good
to be able to use a task from IRAF followed by one from AIPS++
followed by one from IDL if they don't know that these are the
appropriate tasks to use or if they do know about the tasks but
they end up using one of them in an invalid regime.
Keith Arnaud
Lab. for High Energy Astrophysics
NASA/GSFC