[Simh] Way out idea for simh

[Paul Koning]

> On Apr 20, 2016, at 2:01 PM, Sampsa Laine <sampsa at mac.com> wrote:
> 
> 
>> On 20 Apr 2016, at 20:45, Ken Cornetet <Ken.Cornetet at kimballelectronics.com> wrote:
>> 
>> Other than the OS on the old Atari 800 family of computers, I don’t know of any OS that supports a device to which you can supply a file name and then read or write data.
>>  
>> Most OSes view disk devices as a collection of blocks.
> 
> You’re missing my point - the guest OS would not be mounting this as a block device, but would have some kind of file transfer utility to talk to the host OS’s file system.

Let me flesh this out a bit, I think I understand your idea and it's a pretty straightforward one.  It vaguely looks like a pseudo device implementation of the GDB remote stub file access setup.  Or, I suppose, vaguely like FTP requests; they two are the same thing at a sufficiently superficial level.  :-)

Consider a new type of device exposed to the guest software.  You can send it commands: read a file, write a file.  After that command, you read from the device to get the file data, or write to it to send file data.  End of file is an I/O status code (for read) or some special device operation (for write).

From the application point of view this isn't all that different from guest OS file read/write calls, except that (a) it's sequential only I assume, (b) the operations are represented as device operations rather than being handled as OS calls.

What you need for this to work is a way to talk to a raw device.  That means directly, if the OS allows it, or if you don't have one.  Or via a very simple device driver if one is required.  For example, on RT11, or an IBM 1620, you could do the I/O directly.  On RSTS you'd either need a driver (which is a pain) or do it through a sequence of "peek" and "poke" operations (not too bad).

Yes, that seems like a notion that could be interesting.  It would be good to do an existence proof, on some not too difficult machine.  Perhaps a PDP11 or PDP8, with the "direct to the device" approach.

	paul