[Halld-offline] discussion/proposal for making transition to G4

[Shepherd, Matthew]

Hi all,

Context:  this issue on GitHub  https://github.com/JeffersonLab/HDGeant4/issues/89#issuecomment-468412911

There is discussion about a consensus/policy for making a transition to G4 and leaving G3 behind.  The reason is that it will likely take a non-zero amount work to fine tune both G3 and G4 to match data.  It is not clear that tuning both is a good use of resources as we can only use G4 in the future due to DIRC implementation.  Let's spend the extra brain cycles doing analysis (or solving some of the other pressing problems we have).

Mark has suggested using CCDB and variations to manage different tunings, which is an option.  This does not eliminate the work required to determine the different tunings for both simulations.  And we may not get them to the same level of agreement.  And this will leave open the perpetual question for every analysis:  did you use G3 or G4 for the Monte Carlo because the systematic errors may change?

In my opinion, it is sufficient for the software group to adopt a policy and enforce it.  They are a thoughtful bunch and solicit input from many people doing analysis work.  Since only the 2017 data is in active analysis, I'll make the following proposal that you can throw darts at: 

Any analysis using data collected after Spring 2017 that needs simulation, must use G4 for simulation.  G3 is officially supported only up through the Spring 2017 run.  It is the responsibility of the individual analyst to quantify the systematic uncertainties on any quantity, e.g. efficiency, obtained from G3 simulation.

If this policy is adopted, I would propose also a modification to mcsmear that causes it to abort when trying to smear G3 output (if it is possible to distinguish from G4) from any run number beyond Spring 2017.  You can have a flag to override this if you want, but it shouldn't be possible for a person to keep running the same code just fine and then all of a sudden get incorrect simulation.

Matt

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4039 bytes
Desc: not available
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20190301/045cd04d/attachment.p7s>