in linux, it is very possible to mount anything as a storage device. cool.

but how to do in re-windows, and of course, in hobby oses.

i have a "working" drive virtualization in fool, but it uses the MS-DOS drive services (not a big problem).
theorically, MS-DOS would then support the virtual drive transparentlly, i did'nt tested it, but there are many cases where it can't work, for example, using direct hardware driver, there, the virtual drive would no longer exist.

and about hobby oses, i just wonder if the architecture think about that problem.

the solution is to implement the drive capability in 3 layers.
the hardware layer (I/O ports, DMA, FILE, RAM, network ...)
the translation layer (localise, translation, paging, ...)
the drive service layer (READ, WRITE,...)

then, the user application (file system, direct to drive, ...) will call the drive service layer, and don't care about the hardware and translation layers.

it would be cool to design a virtual drive architecture for fasm that any user would follow to be compatible with implementations from other coders, and of course, don't use the existing bloated ways, design a new way, a new standard ( http://xkcd.com/927/ ), a new yeahhh, for assembly programing using fasm