> the receiving side silently casted to 32-bit floats, losing data (as the RFC permits) Had to look that up. Was under the impression the rfc only specified th

> the receiving side silently casted to 32-bit floats, losing data (as the RFC permits) Had to look that up. Was under the impression the rfc only specified the syntax. But here it is > This specification allows implementations to set limits on the range and precision of numbers accepted. I don’t think the implication is that it must be done by silently loosing precision though. A loud error would fit that description just fine. So in this case I would blame the implementation not the spec.