We must consider another directory structure so that we can re-use previous pipeline steps, eg with a hash of the parameters. Also the lib or binary tool version should be at least stored.
Actually for the viewer we rely on openMVG being used for the reconstruction (say a sfm_data.json formatted for openMVG), and on openMVG2PMVS undistorted images.
In the job.config object read from doxel-workerflow script we can now specify the reconstruction pipeline steps we want to follow when running bin/doxel-workerflow, at the time I write this openMVG + CMVS/PMVS or openMVG + SMVS, but everything else can be added (executables must be run in docker containers, eg: https://github.com/doxel/openmvg-docker and https://github.com/doxel/cmvspmvs-docker or https://github.com/doxel/mve-docker)