FreeVMS

From: Brian Wheeler <bdwheele_at_indiana.edu>
Date: Mon Jan 7 08:29:57 2002

On Mon, 2002-01-07 at 01:28, Ron Hudson wrote:
> Remember the below comments are based on a rusty view of
> vms v3.x (before command histories)
>
> I have "PC DCL", it's a DOS shell that looks / acts like DCL.
> It's fun while it lasts but I miss some things about VMS/DCL
> - Authorize
> - Privilidges
> - the "version number" nnnnnnnn.eee;vvv
> - The other things I never learned about as an operator
>
>
> I think the tack of taking one of the Linux/netbsd... etc
> and putting on a DCL shell is not what is wanted.
>
> An X86 machine would never be able to run VMS binaries, but
> I don't think *nix is close enough to VMS to perform that
> feat.

Well, I don't believe that even Itanium has the 4 level protection
structure that the VAX/Alpha has, but the Compaq engineers are porting
VMS to it anyway.

>
> The rights situation is all different. The filesystem dosn't
> have version numbers, and I think most of them keep diferent
> dates (not the same dates as VMS).
>

Why not use a unix kernel as simply a layer for device drivers which
provides a common interface to the hardware? Sure there might have to
be some funky workarounds for things like shared memory, and process
structure, but those are implementation details. The AROS project did
this for AmigaDOS: built the libraries as linux shared libraries and
executables. When that was working well enough, they wrote a simple
kernel which provided enough services for them to load the libraries and
executables. There is a caveat compared to AROS: AmigaDOS is a shared
address-space OS, whereas VMS has separate address spaces for each
program. This might be a problem, though I believe there has been a
unix on unix implementation (where each program got its own process
space).

A similar approach might work for VMS. Implement a 'kernel' as a unix
program, and work from there. Things like filesystems and privs would
get handled by this program (or others linked to it) and could be tested
one at a time until a base system is available, at which point it could
be implemented on other hardware as standalone.

Brian Wheeler
bdwheele_at_indiana.edu
Received on Mon Jan 07 2002 - 08:29:57 GMT