M.I.T.'s Information Services and Technology organization has released the source code of MULTICS, a decades-old OS and important forebear of modern day operating systems. Although there are no systems in operation today that could run MULTICS, the code release may serve as a fascinating research tool for computer scientists …
"There are no systems in operation today that could run MULTICS" ... no, really?
Yes, in the sense of launching the existing binaries, but a 36-bit word is no longer a major hurdle, the bulk of the code was in PL/1, a fairly regular 3rd generation language ... and the guys who played with it originally are probably just retiring and in need of something to keep themselves busy.
So how long before a new project shows up on Sourceforge?
Typical MIT overclaiming :-)
"It was also the first to use the modern standard of per-process stacks in the kernal, with a separate stack for each security ring."
The Burroughs B5000, which shipped in 1961, predated Multics.
The Unics (Unix) name in-joke: it is pronounced "eunuchs", because it's a cut-down version of Multics.
Multics on a chip anyone?
Yeah! Now work can continue on the Multics on a chip!
I've yet to see anything that comes close to matching the security that Multics had. I also miss the flexibility that the Multics ACL used to give for access control. That was how it should be done, not the pale imitations we have now.
Of course, these rose-tinted hindsight glasses might be at fault. I don't think so though.
@Rippy: the main issue would probably be replicating the necessary hardware ring structure required for the ring gating to work correctly. Although the necessary hardware might be doable on a modern FPGA at this point... hmmm....
Burroughs did many things first...
but protection rings was not one of them. Calls to the Master Control Program (MCP) and other protected subsystems took place on the _same_ stack, but with hardware-enforced descriptor-based protection. However, despite its many innovations (e.g., MCP was written in Algol), essentially no subsequent non-Burroughs systems of any significance derived much from those pioneering efforts. Its closest architectural heir might be the Intel 432(!)
OS historians may argue about the significance of specific developments, but many core features of today's operating systems--processes, paging, uniform virtual memory, programming environment, hierarchical file systems, ACLs, I/O streams, security approach--derive directly from the 1964-1965 Multics design.
An interesting paper...
Thirty Years Later: Lessons from the Multics Security Evaluation
Re: yeah, right (hardware replication)
Virtualisation! Can't wait for someone to release a suitable emulator so I can have a play.
You lot are geeks! *coat please*
"Although there are no systems in operation today that could run MULTICS, the code release may serve as a fascinating research tool for computer scientists and academia."
I give it six months, maybe a year at the outside, for someone to get Multics up and running in some form, be it 'natively' or as an emulator. You really don't want to go making throwaway comments like this - someone might take you seriously :-)
Phoenix Systems for Disaster Recovery .....
You will find the current manifestation/reincarnation/strain of MULTICS in the wild, within Quantum Leaping ahead into the Near Future .....By amanfromMars Posted Tuesday 13th November 2007 14:14 GMT available for ESPecial Relativity Comparison and Peer Review here ..... http://comments.theregister.co.uk/2007/11/13/marvel_digital_comics/
<<<< Multics on a chip anyone?
By yeah, right.
Posted Wednesday 14th November 2007 05:22 GMT
Yeah! Now work can continue on the Multics on a chip! >>>>
That would be a Mobile Communicating Chip, yeah, right. 4HQ2HQ Head2Head XXXXChanges.
It is as well for all to know and remember, as I'm sure all will realise if they only think, that for as much as we may have and know with regards to Technology and ITs Communication and Use, that which is being dDeveloped with IT, Stealthily in Secure Private Labs and now Virtually BetaTesting ITself in and for the Myriad Realised Establishment Operating Systems which currently exist in such an ad hoc global environment as is Today and which does not provide an Open Secret Transparent Connect, will by the very Nature of Growth in Binary Processing Power, Dwarf and Eclipse/Simply Replace Existing Systems with New Memory and Content to Play with, thus to maintain Stability in the Status Quo even as IT Replaces it with Advanced IntelAIgently Designed Systems Hosting Virtual Reality.
And any doubts which you may have about that are yours to share and mine to dispel for I do not doubt any of the above and suspect that there is also so much more to be added, if only one is so sure as to have diligently betatested the System Parameters/Program Defaults on oneself.
Please please please someone write an emulator this will run on!
Greenberg's Emacs is the finest
Bernie Greenberg had the insight to write Emacs in LISP: Stallman's original was a set of TECO macros, as its name implies. Greenberg saw that having an editor written in an interpreted language was the way forward, and although a few parts got written in PL/1 (and, I think ALM) that vast majority of it, including critically the redisplay loop, were in LISP. Redisplay is nothing like as big a deal now, but when a 9600 baud link was fast it was a very big deal indeed.
But Greenberg wrote a lot of Emacs in Latin (buffer-est-delenda-p, etc) or cod German/Yiddish (der-wahrer-mark was what was set by ^@). You can see the handover from Latin to English at the top of http://web.mit.edu/multics-history/source/ldd_listings/unb_2/e_redisplay_.list.
you are correct - remember these are the kind of people that designed a TCP/IP network that runs using carrier pigeons :-)
Actually... it was quite interesting. Multics was before my time and since I've never learnt the "history of computing" in any detail, this was all new to me.
I think the world could use an OS designed with security as _the_ goal not a bolt on driven predominantly by marketing. Surely it would be better to run a Windows emulator (that only gets to play in its own little sandbox) on such a security concious system than attempt to emulate that secure system on something that is inherently insecure?* Especially if the x86 architecture can support these security enhancements.
* and before the Vista/*nix/Mac zealots get all flamey/smug, read the article, it basically states that none of the current generation of OSs was really designed specifically for security - for all of them security is an add-on; an additional layer of complexity (therefore more likely to suffer from flaws).
Could be useful in patent disputes
I'd like to see how many software patents could be undone by the release of this code. How many "novel" and "innovative" "inventions" that can be proven to be neither novel or innovative.
Books were written by Elliott Organick about both the Multics system and the Burroughs B6700 system, an evolutionary development of the B5000:
"The Multics System: An Examination of Its Structure"
"Computer System Organization: The B5700/B6700 Series"
I'm shocked at the price asked for these: gave my own away some years ago, alas!
Bernie Greenberg and his use of forign-language terms and expressions
@AC, "9600 bauds", you must be joking! Started using Multics Emacs at 300 bps and moved on to 1200 bps.
Back in 1996, I asked Bernie Greenberg about this. I just posted his reply at http://gizmonaut.net/blog/software/multics_source_reveals_le_jetteur_des_gazongues.html
Multics has been the wild for three decades...
Multics was developed with US government money, so clever folks at Honeywell, Prime and Stratus got the code for (virtually) free by filing a Freedom of Information act request. Stratus' VOS is a Multic derivative that is still for sale and used on mission critical systems such as the defense center at Cheyenne Mountain, CO not to mention stock ticker plants.