Attending: Patricia, Liz, Emil, Ben, Pere, Graeme, Guilio, Michel, David

• Timescale is reasonable, but people do need to find time to write, hopefully together

New things added to the scope:

• Validation
  • Can vary experiment to experiment
  • There is scope for common tools in this area
    • Machine learning (anomaly detection is potentially interesting)
  • Statistics of validation histograms are important
• Interaction with deployment and operations
  • Document the most popular solutions, but stay generic
• Licensing (take ownership of that here)

Document Discussion

• Make software more modular? Brings problems of assembly and combinatorics.
  • Industry is in much more control of the stack than "we" are
• Is it possible to put everything into the one repository? Probably not a good solution.
• Are our big products usable in all cases?
• Two challenges, at least
  • Integrate the work of others in a particular experiment
    • Consistent and reproducible stack
  • Integrate the work of dispersed development communities even within an experiment
• How can we actually monitor our performance and measure the success in meeting challenges?
  • Patricia is concerned that without this we write a lot of empty words...
  • Should strive to identify areas where HEP needs to do work
• Integrating with non-HEP software is a challenge
• We are not trying to address the whole of data preservation, but in so far as one can write code that's more future proofed we should say something
• How to move away from software components that reached their end of life? Internal and external.
  • e.g. Pythia6 vs. Pythia8

Suggestion (Pere) to rework the document in terms of challenges and solutions. May bring more cohernece to the chapter.