Scientific software has a survival problem that ordinary software does not: the physics inside it was correct when it was written, has been checked by people who are now retired, and is encoded in a language most working programmers have never opened.
This is a guided tour of a real stellar evolution code — the numerics, the layout, the conventions that look like mistakes until you understand what they are protecting. We look at what forty years of careful maintenance buys you, and what it costs to change anything.
The argument, if there is one: the fastest route to being confidently wrong in scientific computing is rewriting something you have not yet understood.
What people leave with
- Why Fortran persists in numerical work, on the technical merits rather than inertia
- How to read unfamiliar scientific code without breaking it
- What a modern Python layer around an old core can and cannot fix
What the room needs
- Projector or large screen
- Wi-Fi helpful, not essential
- Works well as a lab session
Where it sits
This talk is ringed. The others are links.