On 20 Jun 91 GMT Craig Hubley <·····@gpu.utcs.utoronto.ca> wrote:
>Well, OK you can laugh. I am certainly asking for it. But if you know of
>such a beast, please let me know. I can't imagine that anyone else would
>be interested in this, but I will happily repost any reply I get if there
>are any. Since I'm already over the line, I might as well ask for CLOS
>and DB2 connectivity too.
Lucid Common Lisp is available for IBM/MVS (I think they have a VM
version as well). IBM used to market it as their own, but gave up
when they failed to sell enough. Not surprising, considering the
way they packaged it: you need a PC running Microsoft Windows that
is locally connected to your mainframe, and you have to purchase
licenses depending on the number of users, etc., etc. Lucid may
have eased up on some of those conditions.
Here at Draper, however, we have been running a homegrown MVS Lisp
called "Zil" on our IBM 3090 under MVS/ESA. It's not 100% Common
Lisp, but is compatible, and also has some Maclisp features.
The Maclisp features were needed for DOE-Macsyma, which we ported from
the VAX NIL version - it's arguably the fastest Macsyma in the world
as of now. We also have ported OPS5 and are using it for a number of
in-house expert systems.
Zil runs in foreground using TSO on plain IBM 3270 terminals, or in
batch (that's right, JCL). It doesn't have CLOS, but it does have
FLAVORS, for what that's worth. It doesn't support DB2 (I know that
Lucid's does), but it does support ISPF services and TSO commands.
Debugging is minimal - it has TRACE and it will display the control
stack when an error occurs, and that's about it. For an editor, it uses
the ISPF editor (if ISPF is available).
The garbage collector is a combination of mark-and-sweep and
stop-and-copy, and also utilizes the 3090 Vector Facility. The entire
system exploits MVS/XA, using memory above the 16-megabyte line.
Right now it's for in-house use only, though Boston University had a
test version running on their VPS system as part of a special
arrangement. We have no plans in place for distribution, although
we have given it some thought.
>If you don't know of a mainframe port, a well-connected version for the
>RS/6000 would be a reasonable substitute.
Me too. I got some literature from Franz, but I don't have enough disk
space to do any 30-day evaluations :-( - you might want to try porting
akcl or one of those if you don't want to spend money.
Hope this helps.
- Steve Bacher
- Draper Laboratory