I feel this take misses the picture a bit in terms of the strengths and weaknesses of FOSS vs commercial software. FOSS is great at building tools for common or popular problems, but starts to run into challenges in solving problems that are some combination of niche, unfun, too big, or too hard.
For example, I needed to export thousands of scheduled jobs off of an IBM mainframe and onto a different platform as part of switching to a COTS ERP. Should I task an internal team of developers to write a one off? Would any open source solution exist? There aren’t a ton of zOS open source projects out there, in part because there just aren’t a lot of zOS systems programmers out there. They’ve all been frozen in carbonate to solve the Y3K problem, lol. No, in this case I light a pile of money on fire and pay Computer Associates for their commercial tool which generated an XML file almost a million lines long (just the jobs and scheduling parameters/dependencies). And it just worked, insofar as any ERP migration just works.
Another factor is the time it can take for a FOSS project to mature. No one would try and say that Octave is a 1-1 replacement for Matlab. Indeed, it wasn’t until jupyter notebooks came along in Python that I felt I really had a good Matlab alternative, and even then, some less common packages don’t have a good FOSS alternative in Python. I still remember the first time getting some of the open source convex analysis packages going on Linux. It was a nightmare of dependencies and didn’t have all the capabilities of the commercial solutions, because that type of mathematical software development is really, really hard.
Additionally, commercial software is helpful at supplying services with ongoing costs. E.g., office 365 with OneDrive would require rolling my own NextCloud with libre office or something similar to get anywhere near the functionality I get from a family Microsoft 365 account out of the box.
I’m all for FOSS, but a tool is a tool, and sometimes commercial software fills needs that just aren’t going to realistically attract a developer community. However, my favorite client tools are usually open source and I like being able to pilfer the code for my own projects.
Edit: I also wanted to add that I did commercial software development at one point, and I got to solve some really difficult, deep technical issues. It may sound really lame, but the work I did on search optimization for a commercial tool was really rewarding. Taking something that lots of people had to use and improving the execution time of arbitrary queries from minutes to seconds was a blast and important to the org, and something I was able to take the time to get right. I’ve also just had to bang out some CRUD code too, so of course it varies, but not every commercial code outfit it terrible to work for. And the hardware and OS on tools like Palo Alto firewalls just wouldn’t get made through open source alone.
I feel this take misses the picture a bit in terms of the strengths and weaknesses of FOSS vs commercial software. FOSS is great at building tools for common or popular problems, but starts to run into challenges in solving problems that are some combination of niche, unfun, too big, or too hard.
For example, I needed to export thousands of scheduled jobs off of an IBM mainframe and onto a different platform as part of switching to a COTS ERP. Should I task an internal team of developers to write a one off? Would any open source solution exist? There aren’t a ton of zOS open source projects out there, in part because there just aren’t a lot of zOS systems programmers out there. They’ve all been frozen in carbonate to solve the Y3K problem, lol. No, in this case I light a pile of money on fire and pay Computer Associates for their commercial tool which generated an XML file almost a million lines long (just the jobs and scheduling parameters/dependencies). And it just worked, insofar as any ERP migration just works.
Another factor is the time it can take for a FOSS project to mature. No one would try and say that Octave is a 1-1 replacement for Matlab. Indeed, it wasn’t until jupyter notebooks came along in Python that I felt I really had a good Matlab alternative, and even then, some less common packages don’t have a good FOSS alternative in Python. I still remember the first time getting some of the open source convex analysis packages going on Linux. It was a nightmare of dependencies and didn’t have all the capabilities of the commercial solutions, because that type of mathematical software development is really, really hard.
Additionally, commercial software is helpful at supplying services with ongoing costs. E.g., office 365 with OneDrive would require rolling my own NextCloud with libre office or something similar to get anywhere near the functionality I get from a family Microsoft 365 account out of the box.
I’m all for FOSS, but a tool is a tool, and sometimes commercial software fills needs that just aren’t going to realistically attract a developer community. However, my favorite client tools are usually open source and I like being able to pilfer the code for my own projects.
Edit: I also wanted to add that I did commercial software development at one point, and I got to solve some really difficult, deep technical issues. It may sound really lame, but the work I did on search optimization for a commercial tool was really rewarding. Taking something that lots of people had to use and improving the execution time of arbitrary queries from minutes to seconds was a blast and important to the org, and something I was able to take the time to get right. I’ve also just had to bang out some CRUD code too, so of course it varies, but not every commercial code outfit it terrible to work for. And the hardware and OS on tools like Palo Alto firewalls just wouldn’t get made through open source alone.