Unfortunately, the line between the two is blurry at best. For example, I'm not sure how well this jives with the case of the printer software which provided the catalyst for RMS starting the FS movement in the first place.
See for example, "Predictable Programs in Barcodes":
From the abstract: "...In particular, we consider programs for microwave ovens, which provide a basic open API for controlling cooking times..."
The authors were only able to _simulate_ a microwave oven, presumably because the software on real microwave ovens is closed. A reasonable argument could be made that the proprietary nature of microwave oven software has limited innovation in this market.