Here we collect a list of question to be discussed during the session
-
The benchmark should be usable (and not modifiable) by others (e.g.: hardware vendors)
- In their invitations to tender the sites ask the hardware vendors for benchmark results.
- It is mandatory to grant licenses of the new benchmark suite to the hardware vendors.
- As an example SPEC benchmarks are publicly available (not for free, but its an industry standard benchmark, and the major hardware vendors have purchased their individual license).
- Other benchmarks as whetstone, linpack are publicly available for free but are not replicating HEP performances
-
What is the public license models suggested by CERN?
- We note that some public license models require that not only binaries but also the source code will be distributed if requested!)
- The suite is very flexible and configurable but it is important that the score used to assess the performances of the CPU nodes, both for the procurements and for the pledges will run without modifications.
-
How the license of the included packages affects the license of our product?
- we do not rebuild the experiment software. We just install it (copying from cvmfs) and distribute via containers
- In addition we install software from public repositories, via yum install (from CERN, EPEL, WLCG repos) and pip install
-
Copyrights
- The new benchmark tool is a composite work of several authors from several sites.
- What are the best practices to acknowledge the work done by each contributor?
- Is a contributor file in the repo enough?
- shall each file contain the header with the author list? (Hopefully not)
- Shall each contributor sign a document giving the right to CERN to distribute this package? Is that implicit if we put a license?
- Is Zenodo an alternative to claim Copyrights? or is it in addition?
