The ili2pg workflow offers some benefits (parsing/writing already implemented), but also many drawbacks (speed, a 3rd datamodel definition with different constraints, java dependency, difficulty to deal with SIA405 VS VSA-DSS that don't properly inherit from each other, etc.).
We could consider again writing/reading the .xtf directly, as these data formats are simple and well documented, and keep ili toolkit only for pre/post-processing (ilivalidator, translating xtfs to french, normalizing data format, etc.)
The ili2pg workflow offers some benefits (parsing/writing already implemented), but also many drawbacks (speed, a 3rd datamodel definition with different constraints, java dependency, difficulty to deal with SIA405 VS VSA-DSS that don't properly inherit from each other, etc.).
We could consider again writing/reading the .xtf directly, as these data formats are simple and well documented, and keep ili toolkit only for pre/post-processing (ilivalidator, translating xtfs to french, normalizing data format, etc.)