This means adding experimental data to files without (externally) defined semantics.
This "temporary persistence" should allow experimenting with various properties/options while not being burdened with the nightmare of version-managing them.
[ ] Engine versions that can load other (old) versions should ignore the exp. data of those other versions by default.
[ ] In addition to the existing version-difference warnings, the presence -- and rejection -- of such exp. data should also be warned about.
New versions might also opt to reuse old exp. data from existing models, even with totally unrelated new semantics (if the underlying on-disk format is suitable).
[ ] Some sort of custom attribute or (comment? ID pre-/postfix?) markup would be needed to semi-automate it.
(I'm worried about just committing to FlatBuffers (#490), as that also relies on double bookkeeping, which I will surely be unable to keep in sync.)
Follow-up of #361
This means adding experimental data to files without (externally) defined semantics. This "temporary persistence" should allow experimenting with various properties/options while not being burdened with the nightmare of version-managing them.
New versions might also opt to reuse old exp. data from existing models, even with totally unrelated new semantics (if the underlying on-disk format is suitable).