Indeed. My greatest use of Open Source, freeware, shareware and other kinds of "free" software is "what if"-type questions. They would be difficult to answer if all that existed were paid-for commercial solutions that you were then tied into.
Do we need Smoothwall in our large school? Hold on, let me bash out a squid + DansGuardian + iptables setup on an old office machine - look, it does roughly this. Great, should we buy the "commercial" product or is this more-than-enough for what we need (and I usually get both answers over time, depending on where I am)? Actually had one school use my box for 5 years rather than pay Smoothwall nearly a grand a year for updates.
Whoops, we're out of MS licenses and we bought a load of netbooks - there you go, have LibreOffice. While you're there, tell me what's wrong with it and why we couldn't just use that everywhere. Nobody ever came up with an answer to that, which really makes me question why we pay MS for Office.
My last one was digital signage. The school I work for had Powerpoints exported to MP4, then put onto a USB stick and plugged into an LG TV with looping turned on. Looked horrible but did the job. They knew it was the bare-bones and were looking for an all-in solution.
I put in a Xibo box as a test and asked if that was closer to what they wanted. Overnight, the LG TV become attached to a PC running Xibo Client. We've tested it running over RDP from a VM and even off a Raspberry Pi. It's bridged the gap between "an old TV showing something" and "stupendously expensive site-wide digital signage system" nicely. And in fact will probably be as far as we go. If we end up having ten displays showing more than 3 or 4 different schedules, I'll be amazed and it will indeed be time to move to a more commercially-supported package. But for now? A £100 TV and £25 for a RPi box with appropriate cabling. Seems to do the trick quite nicely.
We were going to buy a helpdesk system (don't quite know why). Stuck GLPI on, nobody's ever complained and I've been using GLPI for nearly 10 years in various places.
The beauty of open-source stuff is that you can prototype for free, find out whether there is some element that you will NEED to pay for (i.e. better customisability, more scalability, commercial support, etc.) and not worry about the licence interfering at any point. When you throw it all out, or push a working system into wider deployment, the licensing doesn't really affect you. The only point is does affect you is when you try to commercialise it yourself.
My first reaction upon being asked to do something is "Can I find a bit of free/open software that will do that?". If I can, then we can judge our real needs and requirement. If I can't, nothing lost - and it probably is something that takes a lot of commercial backing to make viable, but at least I know that.
Especially in schools, some bits of free/open software are ubiquitous precisely because they are "good enough" - GIMP, Irfanview, Audacity, Blender, etc.
And when prototyping anything, I tend to find someone's already beaten me to it, and usually by cobbling together open components.
Even the open-source projects, most of the time someone's just cobbled together a lot of other open-source projects and their functionality and just lumped them into one convenient package or written a front-end that relies on dozens of other projects in order to reduce the strain.
If the NSA *AREN'T* using open-source (or some agency-equivalent in a private secure codebase) in a modular manner to build both hardware and software for their "one-off" kinds of devices, then they really need to pull their finger out.