Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Industrial Controller? Windows or Linux?

news.ycombinator.com

Post reply on HN
Print view View original

1–10 of 13 posts

Industrial Controller? Windows or Linux?

#1

Post by martin_e »

I am investigating options for the development of industrial control software that will have to run on a ruggedized network-connected tablet PC. None of this requires hard real-time control. It isn't a machine controller. It's a scheduler for a specific set of applications. It has access to sensors and can issue commands to cause equipment to activate or change states. Also, this isn't a situation where people could get hurt if the software fails. However, it could cause business losses or production halts.

Developing on Windows is easy for a range of reasons. Customers have people who can work with Windows, which means supporting them can be easier. My greatest concern with Windows are unwanted automatic updates. I think Windows IoT edition might bring this under control. Not sure. Also, I need to call a sales rep to understand their pricing/licensing model.

The fear isn't reboots. That's not a problem. It's the potential for an update causing our application to stop working correctly.

The alternatives are, of course, Linux and Android.

At that point we have to consider a multi-platform approach to development.

The one requirement is that it has to be a compiled language. The application-specific source cannot be open for a range of reasons, one of them being potential exposure to liability if source is modified and something goes horribly wrong. That leaves out frameworks that are web-centric JS/Python based.

The options I have considered include:

Qt: I am not comfortable with their licensing models and very slimy process. I watched a video presentation where the presenter broke down the license terms and highlighted deceitful language designed to make you believe you have to pay them money. Not for me.

Uno Platform: C# is fine. I's prefer C++, but that's just me. Being able to use .Net UI components sounds like a potential plus. Don't need much, buttons, sliders, menus, graphs, etc.

.Net MAUI does not support Linux. Other than that, possibly interesting. It does support Android.

Xamarin: No Linux but it does do Android. C# & .Net

Kotlin. Interesting. Java. No.

Wondering if anyone who may have walked this path might have something to share. Thanks.

Re: Industrial Controller? Windows or Linux?

#3

Post by martin_e »

smoldesu wrote: post #2

The GTK library might not be a terrible choice if you end up targeting Linux. It's fast to prototype with (in my experience) and very well-optimized for touch/tablet interfaces. Well worth checking out IMO: https://www.gtk.org/

Didn't know about this. Very interesting. Thanks.

Re: Industrial Controller? Windows or Linux?

#4

Post by yjftsjthsd-h »

I'm not sure I understand your constraints. Accepting at face value that your app has to be wholly non-user-modifiable, what exactly is wrong with web/JS/Python? Putting a web frontend on the system shouldn't let the user control anything interesting since the frontend is just a client over the real HTTP control interface. And rolling Python in with Java et al., why does it have to be a compiled language? Most users aren't exactly going to pop open the app regardless, and if your threat model includes users patching the program and ex. editing out signature checks, it's not like compiled code will protect you from a debugger/decompiler/hex-editor.

(To answer the actual question as posed: If you can use Linux, I think you'd be crazy to use anything else. It gives you far more control, completely removes the risk of any updates you don't want, can allow really slick update strategies (roll forward/back, snapshots, A/B updates, immutable snapshots...), most likely completely throws out licensing fees for the OS. Like, if your team is familiar with Windows, fine, if you need Windows-specific drivers, sure, but otherwise? It's an appliance; Windows was never a good fit.)

(Edit: Also, whichever OS you go with, please decouple your stuff from the OS as much as possible. If it's Windows, make a real installer or portable app. If it's Linux, it depends on your model (do you control the OS, or do customers?) but make a package/docker image/flatpak/static binary/appimage/self-contained install directory/whatever and document the install process. I say this as someone who has had to come in after something was no longer supported; it's so much easier to update the underlying OS when it's clear how you install the app on a fresh OS image)

Re: Industrial Controller? Windows or Linux?

#5

Post by martin_e »

yjftsjthsd-h wrote: post #4

I'm not sure I understand your constraints. Accepting at face value that your app has to be wholly non-user-modifiable, what exactly is wrong with web/JS/Python? Putting a web frontend on the system shouldn't let the user control anything interesting since the frontend is just a client over the real HTTP control interface. And rolling Python in with Java et al., why does it have to be a compiled language? Most users…

The hardware and software are delivered as one unit. Think of buying a fancy smart refrigerator. The user doesn't install anything. They turn it on and use it. In this case we'd pre-integrate the rugged tablet with the application and have it launch in single application (kiosk) mode. In other words, from the user's perspective, they turn it on and use it and it isn't a general purpose computer.

While I definitely like the idea of using one of the many Python or JS frameworks for desktop app development, we just can't have the source be openly visible at the moment. It isn't about end users, it's about others.

Yes, I understand, it is impossible to protect against everything. There are a million ways to mess with any system. I get it.

I don't have an adherence to Windows or Linux. Equally comfortable in either environment. I've been using them both since version 1.0.

Part of my thought experiment at this stage is to try to understand what the best decision might be from more than one perspective. Questions such as what UI libraries might be available in each platform. Or, how difficult might it be to hire competent desktop application programmers in each case. Remember, this isn't about building a website. There are millions of very capable web technology coders. That doesn't qualify them to write code that can interface with and control devices in an industrial installation.

Windows has their IoT version which seems to be designed for this sort of application. I need to get in touch with them to discuss. This isn't a high volume application, maybe 1000 units per year. Not sure how that might fit their licensing scheme.

Cost isn't necessarily an issue. If the Qt commercial licensing approach wasn't such a scary mess, I would not have any reservations about paying for that. The problem is that I have to hire an attorney to actually understand it (if I want to be responsible and informed). I have less than zero desire to hire an attorney to choose a development framework.

Re: Industrial Controller? Windows or Linux?

#6

Post by gtirloni »

If it were me, I would definitely not tie my product's existence to Microsoft licensing.

It seems the most flexible solution is a traditional web stack: Linux for the OS, a backend in whichever language you desire and a web app served to the user (if your product has a display, load the browser in kiosk mode).

Re: Industrial Controller? Windows or Linux?

#8

Post by martin_e »

Lariscus wrote: post #7

You might also want to look at AvaloniaUI[0] for a cross platform .NET GUI library. It is similar to WPF but much nicer to work with. [0] https://avaloniaui.net/

That is also very interesting. Didn't know about it. Amazing how these things don't easily surface when you search. Thanks.

Re: Industrial Controller? Windows or Linux?

#9

Post by jononor »

Hardware (and sourcing) is harder than software. So I would heavily weight which OS is supported for the kind of devices you need, from reliable manufacturers with long support times. Always have multiple sources for the type of hardware used.

You also need to consider who is responsible for monitoring, maintaining and updating the devices OS and software, and whether it is a pull or push type update model. If it is you, then the fact that customers IT department knows Windows is not a benefit (on the contrary, they might cause problems).

When it comes to language and GUI framework, pick based on availability of skills. Having people that know X is more important than what X is. I would try to find a way to solve the modification issue which does not limit your choice of tech. Not being able to use JavaScript/web for UI is a huge limitation wrt access to skills. Tamper protect/detection using cryptography.

Post reply on HN