Changed serialization breaks things about nalgebra

[Andlon]

Agree that breaking changes to serialization formats should not happen in patch releases, unfortunate that this slipped through. I can't personally speak to the necessity of the change.

All that said, I think if you want stable serialization, you should perhaps rather re-implement the structs containing your data types using guaranteed-to-be-stable types from std like Vec. This way you can guarantee stability regardless of what nalgebra or other dependencies do in the future. Otherwise you're likely to run into similar problems again at some point.

[Ralith]

Yeah, while this is a significant optimization/bugfix, it definitely should've been held for the next major release.

[oskargustafsson]

Thanks for the replies, you can close this issue. I just wanted to highlight my situation, as I think others might end up in the same.

I will take my chances going forward, serializing Nalgebra types directly. There is significant work associated with maintaining a whole set of types, just for serialization.