Doesn't that mountlist use the device-based handler? My change didn't effect the name of that. It seems that you're addressing the issue of kernel-based vs file-based handlers, whereas my changes address the issue of device-based vs packet-based handlers.
There should not be any difference between resident and disk-based handlers. If the handler is resident, the resident copy is used. When FileSystem.resource is implemented, things automatically go this way (Mount program supports it). Currently they also go this way, just raw packet-based handler can't be made resident (well, in common case). And genmodule can't use dash as separator. I don't want to touch genmodule there, i'd prefer to implement FileSystem.resource and get rid of own IOFS wrappers in CDVDFS and SFS completely.
Dashes can already be used for a handler if it is both file-based and packet-based.
I know. But why keep disk-based handler if you already have it in the kernel? Why duplicate things?
I've found that the build system doesn't cope well with building both variants in the same dir at the same time. This makes sense to me since the object files may vary between the two because of different #defines etc.
I know. But there's no sense to build both versions at the same time. I once built IOFS versions on hosted for testing. I just decided to package "naked" versions because they are smaller. And to prove that it's possible to use these handlers in original form.