19 Nov
2017
19 Nov
'17
4:53 a.m.
On Sat, Nov 18, 2017 at 4:07 PM, Will Senn <will.senn(a)gmail.com> wrote:
> I thought disk (RK05) and tape (magtape) blocks were different...
For simh they are, but not once UNIX sees them.
Physically 7/9-tapes were variable formatt
ed and could have multiple 'files' on them. UNIX reveals all of this to
user (as do most OSs), so you need put in the simh 'virtual' tape format
support for the size of the 'blocks' and all of the extra things that the
HW supports.
But after the simh 'mounts' the 'virtual tape file' on the host when
it
reads the 'tape', simh strips the meta-data out and presents on the blocks
to the OS. Or on write, simh takes the raw blocks, adds the simulated
metadata and writes that to host file system as a 'virtual tape file.'
In the old days disks physically could also be different formats. But
the 'controller' was used to format the disk. Each disk block included
metadata that the controller used. On DEC (and most other systems of the
day), the disk controller had some way to set this up, usually with the
diagnostic system. The OS saw the disk after formatting (as we do now).
The diagnostics would have decided how big a block was etc... DEC
standardized on 512 bytes per block.
simh could have taken the approach like disks, and then 'virtual disks'
would need the meta data; but could have supported all sorts of file
formats (like Apollo's and Xerox's). But the simulated disk controller
would then need to handle the meta data.
Since, most OSs just looked at disk as 'block streams' simh only needs to
provide for the OS to work properly, is map a UNIX file of bytes into 512
byte blocks. This works for most OSs. As I said, it will not work for
Aegis or any of the Xerox systems which put some of what the OS normally
did in the microcode of the disk controller.
Clem