In a rather long thread, and later wikipedia, I learned a lot about
Lisp Machines and have just recently heard about the TI Explorer
systems. For the initial Lisp Machines, they were single user (but
multi-tasking) systems.
Did any of the LMs have a lisp-based multi-user OS? If not, does that
say something fundamental about the nature of lisp?
--
Everyman has three hearts;
one to show the world, one to show friends, and one only he knows.
GP lisper <········@CloudDancer.com> writes:
> In a rather long thread, and later wikipedia, I learned a lot about
> Lisp Machines and have just recently heard about the TI Explorer
> systems. For the initial Lisp Machines, they were single user (but
> multi-tasking) systems.
>
> Did any of the LMs have a lisp-based multi-user OS?
When used as network (eg. file) servers, Symbolics Genera supported
multiple user sessions. The file system, in particular, had robust
security. Any arbitrary networked application could be just a secure.
But the hardware on Lisp Machines consisted of a single user console,
and there was nothing that seperated console programs.
> If not, does that say something fundamental about the nature of lisp?
It would have been easy to support multiple users.
I estimate think it would haven taken a couple of weeks to bring
up multi-user consol seperation functionality in Symbolics Genera.
But if you wanted fairly strong security, enforced by hardware,
that would be a more involved project, but the resulting system
would be much more secure than the operating systems we have today.
But that's just not what those machines were intended for;
people wanted the whole machine just to themselves.
(And of course we wanted to sell one to each person.)
On Tue, 01 Mar 2005 02:08:49 GMT, <······@news.dtpq.com> wrote:
> GP lisper <········@CloudDancer.com> writes:
>
>> Did any of the LMs have a lisp-based multi-user OS?
>> If not, does that say something fundamental about the nature of lisp?
>
> It would have been easy to support multiple users.
>
> I estimate think it would haven taken a couple of weeks to bring
> up multi-user consol seperation functionality in Symbolics Genera.
> But if you wanted fairly strong security, enforced by hardware,
> that would be a more involved project, but the resulting system
> would be much more secure than the operating systems we have today.
>
> But that's just not what those machines were intended for;
> people wanted the whole machine just to themselves.
> (And of course we wanted to sell one to each person.)
Well, I would have marketed LMs differently, but that is the past.
Thanks, it's good to know that as long ago as the initial LMs,
multi-user was possible. Maybe LMs will come back again someday.
--
Everyman has three hearts;
one to show the world, one to show friends, and one only he knows.
Martha Stewart called it a Good Thing when GP lisper <········@CloudDancer.com> wrote:
> In a rather long thread, and later wikipedia, I learned a lot about
> Lisp Machines and have just recently heard about the TI Explorer
> systems. For the initial Lisp Machines, they were single user (but
> multi-tasking) systems.
>
> Did any of the LMs have a lisp-based multi-user OS? If not, does
> that say something fundamental about the nature of lisp?
Most versions of Unix are implemented in C, and are multi-user
systems, whereas most versions of Lisp OSes were single user systems.
Does that say something "fundamental" about the difference between C
and Lisp?
Nonsense.
The _real_ difference is that the people that wrote C had recently
been working on Multics, a multi-user system and built C as a systems
language for another multi-user system, namely Unix.
In contrast, the people building Lisp Machines were building something
involving a highly interactive REPL model, which, when you have
limited hardware resources, encourages a single user system model.
There are aspects to the LM that somewhat encourage there to be a
single "view" of things, but the lack of multiuser variations is FAR
more an indication that the users weren't that interested in that than
that it would be impossible to implement.
--
output = ("cbbrowne" ·@" "gmail.com")
http://linuxfinances.info/info/internet.html
What's another word for synonym?
Christopher Browne <········@acm.org> writes:
> Martha Stewart called it a Good Thing when GP lisper <········@CloudDancer.com> wrote:
> > In a rather long thread, and later wikipedia, I learned a lot about
> > Lisp Machines and have just recently heard about the TI Explorer
> > systems. For the initial Lisp Machines, they were single user (but
> > multi-tasking) systems.
> >
> > Did any of the LMs have a lisp-based multi-user OS? If not, does
> > that say something fundamental about the nature of lisp?
>
> Most versions of Unix are implemented in C, and are multi-user
> systems, whereas most versions of Lisp OSes were single user systems.
>
> Does that say something "fundamental" about the difference between C
> and Lisp?
>
> Nonsense.
>
> The _real_ difference is that the people that wrote C had recently
> been working on Multics, a multi-user system and built C as a systems
> language for another multi-user system, namely Unix.
>
> In contrast, the people building Lisp Machines were building something
> involving a highly interactive REPL model, which, when you have
> limited hardware resources, encourages a single user system model.
No.
Actually, the background of the people who built the Lisp Machines
was the ITS PDP-10 timesharing system, and Multics.
It was a subset of the very same hackers who did the LispM.
(They were of course also familiar with other mainframe and mini
and micro systems, and with Unix and other toy systems of the day.
And of course all the ideas borrowed from Xerox.)
Lisp Machines were not multiple-user machines because there was no
need to share the resources -- each user could afford a machine
that would at least surpass their own slice of the mainframe.
In fact, the resources available to one Lisp Machine user
far surpassed the total resources available on an entire
timesharing system. The reason that Lisp Machine was created
was because we had very much outgrown the capacity of those
traditional machines.
In most ways, the Lisp Machine had more hardware resources
thnn an entire mainframe: huge address space, far more memory,
and larger and faster disks. (I think the CPU on the LispM was
faster than a Honeywell 6080, even in the early machines;
certainly by the mid and late 1980s there was no comparison.)
Really the only place that the large systems beat out the LispM
was in support for lots of teletypes, line printers, card readers,
and magnetic tape. The LispM instead offered a high-resolution
bitmapped display console with a mouse (and you could have more
than one display). The Lisp Machine was for writing programs
that could simply not be run on the "larger" computers.
The idea that resources were too limited to provide multi-user
service suggests to me that people are not familiar with the
relatively limited capabilities of the other machines of the era,
or else they misunderstand what the LispM hardware was like.
You're comparing "large" computers using RTL with a top
speed of around 1.2 MIPS against custom TTL, gate-array,
and VLSI Lisp Machines. The early Lisp Machine was also
unlike like the other microprocessors of the day, which
were 8-bit CPUs (or even later, the MC 68000 machines).
Consider that many years later, when systems like SUN workstations
came along and were eventually fast enough, and people ran similar
Lisp programs (often ported from the LispM) on "timesharing" systems
like Unix, those conventional machines had to be specially configured,
and did not usually have any horespower left over to actually support
anything besides the single user. A single Lisp Machine was also
faster than the larger timesharing machines like an entire VAX 11/780.
The LispM hardware was competitive in price and performance until
very late in the game, by which time the company's infinite management
problems caught up with it and sent everything down the tubes.
(The last ditch effort was the port to the DEC Alpha hardware,
but that was after "A.I. Winter" and everything, and after the
game was pretty obviously already over, anyway. The company
filed for bankruptcy for the nth and final time before that
project was even completed, and never emerged.)
To make it into a true secure multi-user timesharing system,
as I already mentioned, the Lisp Machine software (from the microcode
all the way up to the applications) could have been changed without
any difficulty. One might possibly want to make some changes to
the CPU to support this, and definitely to the IO hardware,
but that would still not have been a very big deal.
The point is, again, that Lisp did not in any way limit the
ability of the machine to do timesharing. (It could be well
argued that it would enhance such a capability, which is born
out in the fact that the machines were also multi-user server
machines providing good application level security based on
the underlying features of Lisp, even without any multi-user
protection support in the operating system or hardware.)
> There are aspects to the LM that somewhat encourage there to be a
> single "view" of things, but the lack of multiuser variations is FAR
> more an indication that the users weren't that interested in that than
> that it would be impossible to implement.
That part is correct. They were not multi-user systems because
people simply did not want a multi-user system.