When Microsoft first embraced Docker and containers some years ago, no doubt people thought it was trying to nab some of that “virtualization-lite” action for its own Windows platform--at least I know I did. But with the release of WSL 2, which includes a full Linux kernel in Windows, it looks like this strategy has changed <https://www.theregister.co.uk/2019/06/18/the_future_of_docker_on_windows_is_linux_says_docker/>: The Docker news is of no interest to developers using Windows Containers, but there are not many of them. At Microsoft's Build conference last month, Gabe Monroy, lead program manager for the Azure Container Compute team, was asked whether Windows Containers are for legacy and Linux Containers for new projects. "I think that is a fair description," he said, demonstrating how the company's thinking on the subject has shifted. It seems Microsoft is now pushing the concept of develop on Windows, deploy on Linux. It is clearly hoping that those Linux deployments will happen on its Azure cloud, and so it still profits that way. Also, WSL 2 actually makes for worse Linux-Windows integration compared to WSL 1, which emulated the Linux APIs on top of Windows: Similarly, WSL 2 makes life better for developing Linux applications on Windows, but there are downsides. Running Linux in a VM as opposed to redirecting system calls is better for compatibility but inherently worse for integration. One aspect of this is that in WSL 2 I/O performance to files within the Linux VM is much faster, but I/O performance between Linux and the host is worse. Another is that WSL 2 has no access to serial or USB ports. Seems to me the obvious way to fix this, at least partially, is to flip things round, so instead of virtualizing Linux on top of Windows, you virtualize Windows on top of Linux. Then both your development and deployment platforms become fundamentally Linux-based. How long before that happens?