Speed, porting and certification.
If I am taking ADC samples and need to run DSP on them quickly and update a screen ASAP there always seems to be an order of magnitude (if not more) in speed slower vs bare metal/RTOS. I know there is always going to be a difference here due to the overhead of running something like Meadow but sometimes we just need tiny slices of time of full control to ensure timing constraints are met.
Porting because everything changes daily. Seems to be common where we try to shoehorn in the slowest device possible and then might need to switch to a different MCU in the same line or in some cases use an entirely different manufacturer. This is due to either speed or peripheral constraints/requirements.
For devices that need to have a CSA safety rating software routines are needed to be added to check memory and operation. Would probably be really hard for Meadow to support redundant MCU cores and all of that but there are some applications where we just needed to ensure the memory was not corrupt and some other checks. This required libraries written in low level code to be added.
What would really make me interested in these high level products is if I could write the non-critical stuff in C# and still have control over the lower level stuff.
All of our products are using C/C++ and have life cycles of 30+ years. When a MCU goes end of life we can salvage quite a bit of code due to the structure. Only the low level code needs to be updated to "glue" the logic back to the bare metal.
If starting with the Meadow board that is all there is. No way to add in or change low level stuff and if there is not enough processing capacity available its back to bare metal / RTOS. When selecting a chip/platform against requirements these high level boards usually are ruled out due to the risk. Nobody wants to choose a board that has a chance of not working causing us to go down the road we would have otherwise.
Bare metal/RTOS/Linux are so universal, hackable and portable that it is hard to justify a closed off solution.
Out of my colleagues I am the most open to adopting hardware like the Netduino/Meadow. I even brought one into work to use for automating testing equipment and testing out new ideas to show off the rapid development. Nobody was interested in investing the time to learn/work with it since the Bare metal/RTOS/Linux is guaranteed to work one way or another. There is also the question what if the Meadow does not make it past a few years? Any other platform we could easily port to another chip but with the Meadow it would be dead in the water.
Makes managers/developers very anxious when starting off with a new product and you know there are definite brick walls that will kill your entire body of work if they are hit.