On 12/23/24 17:29, Van Snyder via cctalk wrote:
On Mon, 2024-12-23 at 18:21 -0500, Rod Bartlett via
cctalk wrote:
One site I used to service (USGS in Reston, VA)
had a split computer
room. The left side was for IBM gear while the right side held a
Honeywell Multics machine
Stuart Feldman, formerly of Bell Labs where he claimed
credit for Make
and the first (and slowest) FORTRAN 77 compiler, told me that Unix was
developed by some MIT students who thought Multics was too big and too
complicated, so they wanted to develop a small single-user system. They
called it Unix.
Now Unix (and Linux) are far bigger and more complicated than Multics.
Linux and
Unix still have along way to go to approach that statement.
Though large code bases, they are nowhere near as complicated as what
Multics achieved.
There is a public system run by Eric S which allows anonymous logins, or
he will assign you a user ID on his system and let you have at it.
The security is the best of any system ever developed.
I don't remember which of the Unix gurus said "Version 7 was an
improvement on all of its successors," but I think maybe Multics was an
improvement on all of its successors.
Not sure what that means. The main quibble I
have with current Linux is
the systemd, which is mostly unnecessary as well as the introduction of
"snap" for the Linux system. The later is an attempt to mitigate a
security problem which creates a number of very bad features worse than
the problem it attempts to solve.
The kernel is large, but most of the "complication" is straightforward
within the discipline available in the kernel structure.
Has anybody gotten Multics to work on Intel?
There is an emulator for the Honeywell 6180 with the Multics hardware
support that runs a tape image and backup released by a three letter
agency, and uses the source from the MIT release of such from Honeywell.
As far as performance it seems to run about as well as ever when hosted
on Intel as it did on the actual hardware.
It also runs quite well on a Raspberry Pi.
Thanks
Jim