On Wednesday 15 July 2015 14:49:54 you wrote: > > Or, better yet, I'd make sure that we don't use floating point, at all. > > Drop support for doubles. Small devices with no FPU would have to carry > > a huge FP emulation library in order to interpret them properly. > > I considered this, however it causes a regression in functionality on > the C++ stack. I'd definitely consider seeing if we could find a way to > better encode the doubles, however I believe at the moment, the > performance/size increase over the JSON is enough to back-pat for the > moment.
I'd rather we removed support for non-integer numbers altogether from the protocol. If someone can show good reason for having a floating-point property for a specific device, we can reconsider. And even if we do that, the property should be separate from the integrals. -- Thiago Macieira - thiago.macieira (AT) intel.com Software Architect - Intel Open Source Technology Center
