Those large collab papers are hilarious. I once ordered a paper through inter library loan. I got an envelope with the paper in, but when I leafed through it I realized that it was about 12 pages with only a title and author names. Some one had not looked through to confirm that all the pages were there before sending it, and I could understand why. Didn't count the names but it must have been several hundreds. Good luck assigning author rank from their order!
The Thing is, in today's physics experiments at that scale, you cannot distinguish the amounts of work. Basically everyone in the collaboration will have a non-zero share. A paper about a specific analysis would not be possible without the people who designed the experiment, build it, maintain it, do the actual data taking, wrote the analysis software, do dev-ops in the data center etc etc etc.
So there is only one fair solution, list everyone in alphabetical order.
And if two LHC experiments publish together, that's > 2000 people.
high energy physics is somewhat unique in this regard, as other fields would consider the LHC the apparatus which is maintained by techs who are NOT authors.
I work in the LHC. I can confirm dev-ops, the very many kinds of engineers, people writing analytic software, storage systems and a long long etc. do NOT get referenced in physics publications.
Only physicists (but then there are hundreds of them)
Where do they draw the line? Is a person with a PhD in physics who writes analytics software credited? Or only someone with an official research program/grant?
It seems that there can't be that many people actively involved in the paper, but everyone needs to be on, e.g. the Higgs boson paper, not referenced in a supporting work on, e.g. a detector, or their career is trashed.
In our lab we always push for people to write up what they are working on, which may or may not be 'original'. Usually it turns out that there is much innovative or original work, which without documentation the experiment could not be repeated/analysis would have been different, etc. And that kind of tech note/report should be referenced by all the experiments that tool/technique, even if they are not co-authors of the derived work. For example, an experimenter uses visualisation system Y in producing report X: X->Y reference. The vis system uses FFTW documented in report Z: Y->Z. The source and binaries and build chain for the expt and analysis are in VM images archived with the data.
