LispOS directly on hardware or on Unix kernel?

Chris Bitmead uid(x22068)
Fri, 02 May 1997 10:33:45 +1000

>>>Most of the "Unix" utilities are pretty useless, IMHO. I'd like
>>>to get rid of Unix as much as possible. It is basically
>>>a model of how one should not do it.
>>You might not like how they work, but what they do often happens to be
>>useful. Put a wrapper around them that you like and don't re-invent
>>the wheel. Take for example gzip, TeX, ghostview, Netscape, and ten
>>million other freely available Unix programs.
>Those are "Unix" programs? People are users of those
>on Macs and PCs.

I think you missed the point. They are C programs implemented on an OS
which is either UNIX or mucher closer to UNIX than LispOS. I don't
think you're suggesting LispOS be built on top of MacOS are you? Well,
then if you want Netscape it will probably be the Linux or BSD version

>I want to get rid of stuff like csh, sh, cp, tar, cpio, named,
>sendmail, nfsd, ... and all it is built on.

Me too, but we can't get rid of tar completely because we need it to
import external stuff. And we're going to go down the track of giving
compatibility for these C/Unix/WhateverOS, then we probably need sh
and cp in order to build legacy applications.

>>How long is it going to
>>take for lisp people to re-implement all of them? Why do you even want
>>to try?
>Why would someone want a new OS? Much of the Unix stuff that is
>concerned with reading random files in exotic formats and
>generating other files in even more exotic formats will go
>Replace the "files everywhere" metapher with "objects everywhere".
>Genera demonstrates how to **integrate** software based
>on access to objects and applying functions on them, rather
>then piping text around.

I agree, you are preaching to the converted. But as of now there is a
lot of stuff out there that just plain works, and let's be honest now,
a lot of it will never be implemented in Lisp. It's quite amazing for
example that there exists a free Postscript viewer at all. I think the
chances of anyone re-writing this in Lisp are pretty much nil.

>>>>the apps I want. LispOS will be in a worse situation. Let's keep the
>>>>ability to run UNIX and X programs.
>For me the ability to run Windows and Mac programs would be much
>more important. Unix is only a small fraction of the market.
>Windows NT already happens to conquer parts of the server market.

Well then you will have to make it run on Windows and Unix. I suspect
a lot of the people on this list are free software fans and will want
to see it running on Linux/BSD at a minimum.

>>>I don't need that. If I want to run Unix, I can use one
>>>of our Slowlaris machines. 
>>Like 99% of the population I've only got one machine of my own, and I
>>don't want to have to reboot every 5 minutes to do real work.
>I have no solution for that. I don't really use Unix on my home machines.
>What should *I* do? I would lose Photoshop, MCL,
>Eudora, NewsWatcher, Acrobat Distiller, FirstClass,
>Claris Organizer, Newton Toolkit, Pointcast, MSIE, ...
>Cool commercial Lisp environments like Macintosh Common Lisp 4.1
>and Lispworks for Windows 4.0.0 are to beat, too. I have them -
>why should *I* use the Lisp OS?

Well that's the difficulty with backwards compatibility. First we
decide if we will support it, and if so, the ones that will be
supported will be determined by whoever wants to put in the effort.

>>Most people these days are talking of micro kernels that will be able
>>to run Windows, Mac, OpenStep, Unix all at the one time. That is the
>>way of the future.
>Which has not been delivered yet. Why? Because it is extremely
>complex to put a current OS on a Microkernel. Ask Apple.
>They failed to deliver. Despite a legion of
>developers. And despite of hundreds of millions of dollars.

By the standards of difficulty in OS development, building an OS on a
micro kernel is extremely easy. Apple DID build an OS on a
micro-kernel. It's called MkLinux and it took them a matter of months
to build.

The failure of the other Apple project is due to some other
reason. Probably it was overly ambitious.

I'm not suggesting we should use a Micro kernel for LispOS. But I do
think that having it run on top of other kernels is a good thing and
the way of the future.

>> An environment that can run any and every software
>>you can throw at it.
>Pretty unrealistic. I can have Mac emulators on PCs, PC emulators
>on Macs. Pretty much unusable for *real* work.

Why so unrealistic? I've heard that the Mac emulator on Linux as well
as WABI on Linux are actually pretty good.

>> A LispOS that can only run Lisp is burying itself
>>in the sand.
>On Genera they ran the X-Windows server. It was written in C.
>On Genera they had TeX. It was *not* rewritten in Lisp.
>Go figure. It was not perfect, not fast, but then - it
>was ten years ago.

Fine. If we can build C apps on this hypothetical LispOS without major
hacking then that is what I'm talking about.

>>>>And IF in 5 years time, LispOS is so successful that you don't need
>>>>UNIX compatibility any more, well then you can purge it out.
>>>Sigh, not earlier?
>>If you can re-implement every useful UNIX thing I've got now in the
>>next 6 months, then be my guest. Good luck!
>I don't need Unix-compatibility (the ability to directly
>run Unix software) on my Windows NT. Why would I need it
>on LispOS?

Well you need compatibility with something.

>What are the benefits of a Lisp OS when you are not using Lisp
>software? That is Lisp applications?

2 things:
- developing new applications (which I hope can be ported back to the
real world)
- A better environment for doing work in.