Subject: | |
From: | |
Reply To: | |
Date: | Fri, 4 Oct 2002 17:14:18 -0400 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Good ! This will be important when we're running software in the
underlying OS (eg. Apache) that needs to communicate or synchronize with
software running under MPE. A very reliable socket protocol that could be
used by both OS's would be useful too.
Regards,
Frank
On Fri, 4 Oct 2002 13:32:15 -0700, Gavin Scott <[log in to unmask]> wrote:
>Frank writes:
>> Currently MPE and Posix processes can access each other's files.
>
>Of course you'll still have Posix files inside the virtual 3000 environment
>just as you do today, but these MPE "Posix" files (really just HFS named
>files) will not (by default) have any interaction with the native host
>environment that the emulator is running in.
>
>> MPE running with Linux should retain that capability at least for
>> bytestream files.
>
>This is actually possible, though it's fairly deep rocket science. We
could
>write an MPE intercept library for the file system intrinsics that would
>redirect file access to native host files (maybe not MAPPED access though)
>based perhaps on the file's path name (creating in effect a "mount point"
>for the underlying system).
>
>Going the other way could be done today using out NFS/iX package on the MPE
>side that would let you mount the MPE file system from the Linux side.
This
>is a good solution as it requires no special software, libraries, etc., on
>the host side.
>
>Other "magic" things we could do (given the resources) include dynamically
>mapping the Linux filesystem to the MPE filesystem at the disk I/O level,
or
>by intercepting MPE at the VSM layer or something like that. The great
>thing about an emulator is that it gives you total command of the laws of
>time and space as seen by the MPE operating system running under it, so we
>can perform all sorts of magical things that would be utterly impossible on
>a "real" 3000.
>
>Gavin
|
|
|