After some discussion on the 'Features for 14' thread, I'd like to state more clearly the case for a build of FileMaker Server for Linux.
It's simple. Linux eliminates the costs of operating system licenses and provides tools for availability, management, and virtualization that are not available on Windows or Mac OS. Lowering the cost of deployment per instance raises the value of deploying multiple instances, and deploying multiple instances requires advanced tools for management and virtualization. It all fits together. And it contributes to better performance, higher availability, and better business continuity, thereby making users happy with their tools, happy with us developers, and happy with FileMaker as a product.
I dream of rsyncing instance images back and forth for updates, replication, backup, and disaster recovery. I want to run Server on ZFS, the best filesystem available for the kind of data integrity critical to database applications. I want to dynamically allocate memory to different instances as needed (or not). I want to move instances from one host to another without even shutting them down. I want developers on my team to deploy new instances for testing or development from the command line, from images ready and standing by, or to give outside developers full access to whatever they need without putting the live solution at risk. I want to be able to spin up backups or fail-over images from off-site or on a laptop when a storm comes. I want to bring an instance to life at a remote site, perform some routine activity, and shut it down until it's needed again, freeing those resources for something else. And I want to do it all on cost-effective high-performance commodity servers with enough resources to host entire solutions out of RAM if I want, without having to stop and deal with OS licensing.
I can do some of that on Mac and some of it on Windows, or I can do all of it on either with hours of work done by myself or a contractor and gobs of money, but if I could just get a binary executable built for the Linux kernel, I wouldn't have to. I would get a whole universe of awesome, right off the shelf.
I don't mean FileMaker Server itself has to go open-source. There are plenty of proprietary, licensed applications that still run on Linux. Rendering software, engineering software, data modeling platforms, and business tools all run on Linux while charging a well-deserved licensing fee. But they're easy to deploy in bulk and manage as a fleet because the operating system is out of the way, where it would be an obstacle with Mac OS or Windows.
With long-term-support releases and broader enterprise adoption than even just a few years ago, the risks to stability, performance, or security of running a production-class service on Linux are negligible. With the benefit of tools and techniques developed to run cloud applications, FileMaker Server could find itself in much larger production roles than are possible now. Access to FileMaker Server by the current crop of devops-style talent—which means access to Server using the tools developed for devops—could keep FileMaker Server relevant in an increasingly virtualized and orchestrated application-services sphere.
But for me, today, let me build a service platform for my users that allows me to provide the resources they need to focus on the work they want to do, without being distracted by "what's the most powerful Mac I can afford to buy, and can I afford to buy a spare" or "Who do I talk to about buying more Windows Server licenses, and how many CPUs am I allowed to have." Server doesn't have to be stuck in the corner on a Windows VM or a Mac Mini. There's no reason it couldn't live and thrive in the same environment as virtually every other service I provide, free of the cost and infrastructure overhead of Windows or Mac OS.
Where FileMaker Server is the right tool for the job, it seems to me that Linux is the right tool for hosting it. Could the reasons not to run it on Linux even be that good?