> I will note that Windows 95 *could* use DOS drivers. I/O > performance suffered horribly since DOS drivers weren't thread safe, > but there was a copy of DOS in the system VM for this purpose, even > if it had nothing to do under normal circumstances.
> But to some degree this is a philosophical debate that will be > present whenever you have a transition from a CPU architecture > without protected memory to one with it and you move incrementally > from a single-tasking OS to an environment where multiple instances > of that single-tasking OS are virtualized alongside each other. At > what point does your protected memory management software cease > being an application running in top of the legacy OS and start being the > actual OS? you probably meant 'OS architecture' (not CPU). let me add this: fromm the moment EMM386 is started, *all* interesting work is chained through it. it's just a policy decision to pass every I/O port access to the real hardware, and all interrupts to the real mode kernel. that's the way FreeDOS and it's EMM386 works. MSDOS worked like this as well , up to and including Windows 3.11 (according to Andrew Schulman, and I hope I cite him right) starting with Windows 3.11 for Workgroups, this changed, and EMM386 would take *some* part of OS work. starting with Win95, *some* would be *most*, and EMM386 renamed to VMM32. yes it's a philosophical debate, and my opinion is that Win95 would allow as much protection as possible, while still allowing DOS like behaviour and drivers as necessary. That still doesn't mean that WIN95 would 'run on top of MSDOS'. and the WinXP command line is probably the best DOS ever implemented, for the vast majority at least. Few users of this mailing list are doing OUT DX,13 these days. all of them could use Windows command line, or linux DosBox. Tom _______________________________________________ Freedos-user mailing list Freedos-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freedos-user
