Yeah, I saw that, this is based on newer storport and can probably build only for newer windows. It looks also very much overengineered and corporatey code 😉 Quite helpful as a reference
There is also another copy https://github.com/JoeHsu092015/Windows-NVMe-driver.git
And archives of the mailing list where they were developing the driver https://lists.openfabrics.org/pipermail/nvmewin/

In the meantime, I tested an NVMe Gen3 SSD via a PCI-to-PCIe adapter with a PEX8112 bridge chip.

To make things interesting, I converted the signal back to PCIe via a PL7C9C based adapter. (PEX1PCI1 from StarTech)

It works, and it's actually better than I expected, as the adapters somehow agreed to work at 66MHz.

I also tested a few GPUs with this combination in this video: https://youtu.be/-GEsW0JjgtU (it's not actually that bad to go from 16GB/s down to 0,2GB/s in some scenarios)

Now that I know NVMe SSDs work in a "PCI" slot, I'll try this combination on some older systems (and worry later about how to make them bootable 😀)

they do work in pci slots but my award bios on my socket 7 tekram doesnt detect pci bridges even after i added 3.3V power ;(
or at least i think it is the bios. maybe the adapter doesnt like 5v signalling either.
anyways the adapters work in my dec alpha personal workstation too and bunch of sgi boxes.

Yeah, it's most likely the BIOS. I did use this PCI-to-PCIe adapter to test a GTX 980 on a Pentium 3/866, and it worked, but my initial goal was to test it on a Socket 7 motherboard. The GTX 980 actually booted, but I couldn't install the drivers no matter what. Then I got this comment:

" @cybercat1531
The reason the GPU behind the bridge on the MVP3 can't be configured is because the BIOS actually doesn't setup the necessary memory type range that high up in the PCI address range for a second bridge device.
It really only sets up the PCI-PCI bridge in the chipset and the PCI-AGP bridge range in the chipset.
But the extra PCI-PCIe bridge addressing with higher memory is not configured for the PEX 8112
Newer boards with newer BIOS do it correctly though as they expect more PCI bridges to be added.
I had to solve this issue in coreboot BIOS as well when experimenting with PCI bridges."