What gets in the way of your Raspberry Pi project?

Raspberry Pi is suitable for both microcontroller applications using the Pico and computing platform applications with the Pi 4.

The Pico is a great product for microcontroller applications and it will be interesting to see if they add to their offerings in this area.

The Pi 4 as a computing platform is very low cost and capable as a general computing platform but I don't use it for this because I always default to using my most  powerful computer (Core i7) for computing tasks. There may come a time when Raspberry Pi's are fast enough for my main computing activities, but it is not here yet.

The Pi makes a great platform to learn Linux or Raspberry Pi OS. This is an area where I need more knowledge to achieve reasonably efficient work flows.

I would use a Raspberry Pi as say a dedicated MQTT server or possibly a LoRa gateway, because it is low power, low cost and could run continuously.

I think right now it is these types of dedicated edge servers that are the "killer app" for Raspberry Pi. More focus on software support in these areas would help more people to add such devices to their infrastructure. A low cost, full featured LoRa gateway is an obvious gap in the industry right now. 

Another pretty good application area is in specialized cameras since they have built-in camera interfaces and communications. Pushing this capability further with more software support will help expand this to a much more mainstream activity. Low cost cameras are and will be used extensively in a myriad of applications.

I have a Pi converted into a LoRaWAN gateway, the cost of the Pi ends up being small compared to the rest of it, mainly because the LoRaWAN gateway chipset is pricey (dual chips from memory, although most LoRaWAN gateways end up using three or more). Add in the cost of making it outdoor-capable, and a good antenna, and it is then approaching perhaps 60% of the cost of a ready-made gateway, which for home use is good, but for commercial use isn't significant because of costs of deploying (amongst other factors), so the hardware cost difference becomes negligible.

I've been investigating that path too. A proper gateway is cheaper than a converted Pi.

Whether it's MQTT or LoRaWAN gateways you would've had to create driver software or utility software to handle the messaging received and transmitted etc.

What I struggle with is working out which software package works best for these sort of applications. If it's Python then how do you protect your script from being hacked etc.

I would have a similar question for say using Raspberry Pi as say a USB Hub. It's trying to work out what drivers can be installed (and where I get those) and then what additional software layers need to be created to handle the USB data etc. to build an application.

BigG said:

Whether it's MQTT or LoRaWAN gateways you would've had to create driver software or utility software to handle the messaging received and transmitted etc.

It is not often needed to write drivers, except if you are connecting custom designs that aren't supported yet.

BigG said:

What I struggle with is working out which software package works best for these sort of applications. If it's Python then how do you protect your script from being hacked etc.

The security to deal with the execution of Python or other (compiled?) languages is similar. 
It's not something to handle off in a comment. But changing the default user password, running all packages under a dedicated account, and only enabling what's needed is a good start.

Microcontroller software is also vulnerable. Security awareness is needed for any design that needs to be secured.

Most people end up using the few open source projects, or reference code from Semtech. However, as you say, this is just one part of it, and how do you protect from being hacked.

Ordinarily (speaking generally, because I don't know of any orgs using Pi as LoRaWAN gateway) when using Linux in such a way to make a product, there's a process to go through, which results in some hardening of the OS, and eliminating/removing services from starting up if they are not used.
It's a full-time job, at one firm that I worked at, they had a team that was responsible for producing a Linux build that could be used for the org's products. That Linux build would prompt the user to enter the IP address, DHCP, time server, admin password and so on, just as if it were an appliance. The user might not even know it is Linux, since they will never see a shell prompt.

Regarding knowing what drivers etc to use, I can only speak from my experience, where (say) if network communication was needed, then middleware was used to do that, i.e. no creating sockets for instance was needed in the application itself, i.e. the code can be layered. Add in code (can be separate programs) to do upgrades, logging, restarting after a crash, etc., and it's a non unsubstantial amount of work to turn a Linux computer into a product, but feasible for a small team provided some of the effort (like preparing the Linux build and installer) can be offloaded to a team shared for several products.

Wow, you've confirmed my fears. It's complicated.

I guessing here but I think this is where the Yocto project (or equivalent) is supposed to help, where you develop an image file for your application on say a PC and then bundle it with all the Linux OS drivers into one massive custom image to then be uploaded onto the embedded flash or SD card.

Wow, you've confirmed my fears. It's complicated.

I guessing here but I think this is where the Yocto project (or equivalent) is supposed to help, where you develop an image file for your application on say a PC and then bundle it with all the Linux OS drivers into one massive custom image to then be uploaded onto the embedded flash or SD card.

I'm trying to support that approach, by giving cross-platform how-to's.
These are a first step, to learn and develop on another platform than the final solution.
Once you've taken that leap of faith (trust me, it isn't a leap in skills - steps are repeatable), the art of adding your design to an OS build suddenly becomes graspable.

But that's not the only path. You can learn to develop for Linux. Or adapt your design to run on Windows and Linux. 

Part of what shabaz describes, I will be doing for an active road test. I will make 2 separate C and C++ applications that will run as a service on a Pi, and that use the network APIs. Feel free to follow along and to comment, ask, critique, rant.

edit: true story!

answer 2  : I think that for an IT guy without previous embedded skills, it's easier to learn and build a meaningful application for a Pi than on a bare metal Arduino.
Let's abstract from A: blinky and B: meshing existing code available on the internet into something that works.