HSF Licensing Meeting, 9 May, 2018
Participants: Carl Rosenfeld, Thomas Kuhr, Graeme Stewart, Stefan
Roiser, Ed Moyse, Benedikt Hegner, Dave Dykstra, Marco Clemencic, Scott
Snyder, Pere Mato, Martin Ritter, Andrea Valassi, Michel Jouvin, Liz
Sexton-Kennedy, Davide Costanzo
attached to Indico.
- Reminder on meeting goals and the license problem: license
compatibility makes it mandatory to have a licensing scheme
compatible with the license of all the dependencies we use. But also
constraints by funding bodies that are not compatible with the
restrictive approach chosen by some dependencies.
- Summary of experiments’ approach -
- Apache 2 is what CMS would like to apply
- Various projects moved from GPL to LGPL or other more permissive
- FastJet staying with copyleft, but now GPLv2 or later rather than
GPLv2, clearing the incompatibility with anything else the GPLv2
and LGPLv2. Still obliges clients to go for GPL too.
- Please join
Working Group mailing list!
- LCG AA - there are a lot of GPLv2 packages still - to be checked
again (ACTION on SFT colleagues).
- Default copyright varies country to country - copyright is an issue
especially for changing license (something more likely to happen
with a copyleft license than with a permissive one).
- Copyright to CERN as experiments are not legal entities.
- License GPLv3 targeted, CERN copyright makes possible to relicense
- Adjusted statues and procedures in experiment.
- Plan to be finally endorsed in collaboration board in June.
- Some initial reservations from 2 countries, but now addressed.
- Keeping independent record of contributions as to make them
- Committed to open source its SW by summer 2018 and would like to
move to Apache2 preferably, GPLv3 as a fallback solution:
copyright ownership by CERN will allow further evolutions. Still
trying to understand what is left preventing Apache 2.
- Open sourcing the SW, license/copyright is only part of the work:
also need to clean the repositories from swearwords and other
inappropriate words/sentences: tools can help but all involve
rewriting the history with impact on many tools…
- Situation for Gaudi affects both LHCb and ATLAS - problem to track
down all of the (c) holders.
- Some dependencies on GSL (but isolated); Gaudi plugin service is
GPLv3, (c) CERN - could relicense this (should be easy as it is
quite in line with the recent move supported by CERN of
- Gaudi (c) could naturally go to CERN, as most of the primary
authors are CERN and the main experiment users are also
- A TF created in 2017: still at the beginning of the work, no target
license selected yet, welcoming collaboration with HSF on this
- The solution of ATLAS and CMS is split licensing more than dual
licensing (each file has only one license, but mixed in the overall
- Started to work on assigning the copyright to one holder: KEK would
be the natural candidate, but not a role it is prepared to have
yet. Discussions with KEK in progress.
- What would the fallback solution be for the copyright holder? This
is not clear.
- Would CERN be in a position to hold Belle II copyright? KEK
answer really is not known. DESY could be a fallback.
- Carl thinks CERN (c) could be ok for the collaboration, but very
unclear if CERN would take this role on.
- CERN are developing Contributor License Agreements (CLAs) that could be
useful models (probably not needed for experiments that have
their collaboration boards).
- External dependencies - there are a few GPL dependencies like GSL.
Pythia is more of an issue.
- We do have some traction with the theory community and want to keep
discussing with them.
- They should feel part of the community and participate in this
- Citations are really important for the theorists.
- We should keep a list of citations that should be used.
- ATLAS and LHCb assigned (c) through the collaboration board.
- We should be a bit pragmatic about contacting authors and copyright
holders, but make reasonable efforts to do so.
- Action: Update LCG stack for licenses.
- Next meeting pre-summer.