Meeting of November 18, 1997

Joel Butler gave an introduction on why we need to start thinking about Physics Analysis software now.

Vicky had asked people to come prepared to discuss the following:

a) The requirements on Analysis Tools for  - functionality, platform,
maintainability, cost, open standards, scaleability, etc. ,?
b) What should a working group or groups be asked to do, and how should
it (or they) be organized to carry out this work?
c) How might we effectively compare and contrast proposals made by
proponents?
d) What other background work or evaluations should be carried out as
input to the working group process - e.g.
 -impartial evaluation of code quality and maintainability of Root,
 -compilation of List of known  PC tools as a starting point?
 -licensing cost investigations?
 -invitations to potential  'vendors' to present their products in depth
and their plans for evolution of the product.

There was a lengthy discussion - some of the points raised were:

• requirements must include having a certain reality/existence to the tool
• must have an API which allows you to generate the underlying data model and to convert our data to the format of the tool if necessary
• modularity is important - might want to use all or part
• people will vote with their feet - but support is a real issue and many people do want guidance - there is just too much stuff out there on the net and the choices overwhelm them
• licensing/licensability must be considered
• must consider the interfaces to all the other choices of products/ways of doing things which we have made
• are there any other options apart from - Root, LHC++, develop our own, find something either buy or find freeware, maintain PAW (backup position).
• we might have to support ZEBRA anyway because GEANT 3 might have a longer life than we wish/imagine. So maintaining PAW might not be such a far-out backup position
• must have access to fitting packages
• how important is it to have an interactive version of the programming language you normally write in, like Comis was for Fortran, and ROOT seemingly has for C++. Most people thought a simple scripting language is more important
• might want to dynamically link in our own code (like the way Python does it)
• meta-requirement - difference between a tool-kit and a system
• need to get some Computing Techniques seminars going on this topic. Perhaps send people out to talk to others with data analysis problems.
• discussion on how group or groups should be organized to look into this

---

Vicky White