> > IMHO, image upgrade over serial with bootloader only is handy during > manufacturing. > Not so much during development (JTAG instead), or when devices are in the > field. > Users would much rather not deal with cables during update process :) > Therefore, I think making update process as fault proof with hands-off > from end user > will be key. > > In the maker and hobbyist market I think 'manufacturing' is the same as 'in the field' and I think Newt is heading (if not already) to products (and projects) there.
So I think Newt should allow a third option in line with Kevin's thinking, which is basically copy Arduino's bootloader approach, i.e. always present, serial only and can flash the entire remaining flash space with the user app. >From my own experience supporting a consumer product where your end user would have to access JLINK or JTAG to unbrick a device due to 'errant user app code' is not a great place to be; also, BLE DFU updating can be clunky and troublesome in an environment like a classroom/workshop with >10-20 people, serial is well known and good enough IMHO. Perhaps in the future a portion of 'high' flash can be used for optional BLE DFU functionality such that the boot loader could look for some magic numbers and entry points and offer it, or more simply allow optionally baking in a Newt standard library for BLE DFU functionality into the application code startup sequence (with a configurable time and/or button check) All the best Wayne
