Two sites: Don Mills, Ontario, Canada; Dorval, Quebec, Canada
Don Mills: 1979
Don Mills: 6 L68, 4 SCU, 12 MB memory, 24 MSU0500
[Wilf Mandel] The Multics project started in Bell Canada as a replacement for GCOS Time Sharing System in the Dorval Datacentre which was ever more heavily used and providing poorer and poorer response time. GTSS was to provide transparency for the GCOS users rather than try to retrain the multitudes. Several people had just become available with the sale of a small datacentre (and its lone VM/370 system) to Nortel. The group included myself, James Chubb and Joe Caruso.
[Julie Jackson Sibert] I was hired by Honeywell Canada to do the benchmark acceptance testing and installation for the two Bell Canada Systems. I installed both systems in the summer of 1979. There are a couple of fun stories mostly about having to be snuck into the sites in panel vans at night and on weekends to avoid the striking and picketing telephone workers. During my Bell Canada summer, I reunited with Richard Barnes, whom I had met when he came to USL on a recruiting trip. He was glad when my benchmark testing provided him with some very solid numbers to back up his compiler improvements. (I'd never seen a grown man jump up and down with such enthusiasm!)
Marketing: Swan Tan, John Morrison, Julie Jackson (Sibert), Andre Allaire, Ron Brown, Doug Cooke, Stan Dylnicki, Wally Halladay, Mil Travnicek, Al Dupuis
Site Analysts: Ron Faraday, Randy Pepin, Jerry Kelly, Alf Burnham, Guy DuMouchel, Mel Wilson, Stewart Putt
Development: Richard Barnes, Bob Mullen, John Bongiovanni
Field Engineering: John Potts, Ernie Reimer
[Joe Caruso] HIS: Roger Parsons' counterpart in HIS during contract negotiations for Multics was a fellow named Norm Kane who shortly afterwards went on to manage a Canadian Tire store I believe. Bob Mullen was one of the developers who tutored us on scheduling. Ricarda McDonald from Calgary joined us for a few weeks when we had a major system upgrade going.
Bell Canada Folks
Don Mills: Mark Pickett, Randy Oyagi, Ivy Lee, Odel Lewis
Dorval: Case Timmer (mgr), Ian Stromberg, Sandy Bartlett, Bill Harris, Bilal Qureshi, Wilf Mandel, James Chubb, Joe Caruso, Firoz Dosani, Tom Parsons, Blair Morlock, Bill Abbott, Real Rousseau, Sultan Somani.
Users: Alan Sawyer, Stan Daurio, Jim Lyle, Mansur Halani, Alnoor Manji, Farhad Budhani
[Joe Caruso] At the Dorval Data Centre - the GCOS Timesharing crew was swinging over to Multics - Sandy Bartlett - the Manager of the team, Bill Harris (Bill was a GCOSian at heart and got back to GCOS when Bell implemented BCRIS on the GCOS platform). There were 2-3 other programmers in Sandy's team - Bilal Qureshi - Deryck Beckles - Dave Perks - who wrote ostf - the Offline Storage Tape Facility - Pierre Charbonneau the GCOS FNP techie who also supported Multics FNPS - he raised honey bees and rode a BMW motorbike), and several operators of course. Stan Dylnicki of Dorval was the original Benchmark guy, did the original GTSS benchmark.
[Joe Caruso] At the Bell Canada Don Mills site - Dale Perry, Bill Cockburn, Inge Main and I believe Fouad Hassan were all computer operators, Peter Donlevy (Sysadmin), Sing Tung - the FNP guy. Customer Service - we had Mark Pickett, Alec Ko, and Patti Ranzan; Roland Mathis was the Manager. And as Wilf pointed out in the systems group - myself, Wilf, James Chubb, Prakash Kamat, Firoz Dosani, Jim Lyle, and later Jerry Kelly.
[Joe Caruso] Sing Tung used the Simplex method to determine the optimal number of lines for 1200/300/110 to configure. We had drastically overconfigured for 110.
[Joe Caruso] Other Toronto - the provisioning group - Bruce White was the District Manager and for the Application Group - Peter Langenberg was the Manager.
[Joe Caruso] Roger Parsons Toronto was involved in the initial contract negotiations for the system; he passed away while on assignment in Venezuela in the early 90s.
[Joe Caruso] Wayne Higgerty was a manager involved in Multics Provisioning when Tom Parsons and I worked for him. Also, Noreen McCarthy moved from Montreal to Toronto for a year and a bit to work in the Multics Provisioning Team before moving back to Montreal. Our District Manager was Bob Gleason.
[Peter Langenberg] We also had a very solid MULTICS time sharing team in Toronto, I was forced to move there in 1977 with the big migration but we had John Willins, Harry Wu, Odel Lewis, Barb Garbol in Toronto and had a great time there writing time sharing applications.
The February 81 Benchmark Exercise
[Joe Caruso] I spent 3 weeks monitoring the GTSS benchmark exercise in Phoenix - Project Pisces run by Deiter Fischer - and Bruce White from Bell. - a lot of people involved in that one from Toronto, Montreal, Phoenix. While the benchmarks were running, we played backgammon on the tubes in the nearby room. Doug Cooke and I used some of the active functions to create a menu where you selected and stored the parameters to be used for the run - number of cpus, memory, scripts to run, number of runs - this ensured that before we started the run we had the right parameters to eliminate bad runs. We once forgot to ensure the cache was turned on - we added it to the menu and we only every had that one failed run over the 3 weeks - during the run of benchmarks for the second half, there were three shifts running the benchmarks. We had boxes of listings that were shipped back to Toronto, not to mention all the boxes of tapes that had been shipped up to rebuild our benchmark environment in Phoenix. BTW we only had one day off in the 3 weeks, the day there was a hardware problem and we didn't have to go in - it rained that day - a Sunday.
The Conversion from GCOS to Multics
[Joe Caruso] The benchmark exercise of Feb 1981 showed that GTSS was not any more efficient than native Multics, so the decision was made to migrate off of GCOS directly to Multics. One of the killers was the use of addnames to identify GCOS file properties on a native multics files. I was based in Toronto and moved from the Systems Group to support the application group downtown as part of the conversion. I designed and implemented a conversion tool to convert GCOS Fortran to Multics Fortran. Coded in PL/1. I used Nassi-Shniederman charts (this was in 81) to document the program logic for the coder. The Dorval application team borrowed the skeleton of our code and built a Basic Language source code converter. The Dorval app team also built a data file converter. The first tool handled about 80% of all the data files - we unfortunately began to discover there were exceptions and Gaetan Biron the larger (there were two programmers with the same name) ended up coding about 3-4 other data file converters. I gave a talk at the Toronto HLSUA on the Conversion from GCOS to Multics. I remember the microphone at the podium kept cutting out. We converted one of the largest applications first, SOFA II and it went pretty smooth.
[Joe Caruso] During the conversion from GCOS to Multics, in Toronto, Jim Sawada was the group Director. He reported through to Bill Tawse. One of the Dorval Managers during the conversion was Jacques Chicoine.
[Joe Caruso] We also had a newsletter during the conversion - Communique - hey notice it works in French and English? My gosh how far we've come - we had a metal plate to get the crisp quality graphics on the cover page. French was Red Logo, English was Blue.
[Joe Caruso] The Toronto Application Group during the conversion included the Provisioning Team - Roger Parsons - Manager, Joe Caruso, Tom Parsons, Blair Morlock, Jyki Pajunen; Wayne Higgerty - App Manager, Mel Walker, Barb Garboll, Fred Wu, Sylvia Eng, Lori Pietras (now with the OPP IT group), Bruce Frazier, Laurel Nolan; our secretary was Tiziana Papi; Jurgen Wulf newly returned from Saudi replaced Roger Parsons. Director was Howard Robb and his boss Bill Tawse - Bill was part of the old school Bell managers who were trained at Cherry Hill NJ when Bell was an ATT sub. Our AVP was Al Smith, who supposedly at one time was speech writer for the Queen of England - never did confirm this. Question - how and why did the conversion exercise end up under the VP of Admin Services? The job got done though. The lesson learned here was put everyone reporting to one manager as far down in the hierarchy as possible - it improves communication and gets decisions made faster. Bill Tawse was the focal point - for the App Development and the Operations (systems, operations, customer service) districts.
Alan Sawyer has provided the story of the SOFA II conversion from GCOS to Multics and Multics to IBM in his Multics bio.
[Wilf Mandel] As an aside, I can recall our management being mystified by the concept of electronic mail, which we accessed on System M in Phoenix. Dial-up analogue 1200 baud on an ASCII terminal was impressive!
[Wilf Mandel] I remember the struggle to get DATAPAC access going as a way of eliminating large modem pools... Thinking about this makes me think of horses and buggies for some reason - broadband internet and wireless were dreams back then. On the other hand BBN were already on the way with the beginnings of the internet with ARPANET. Recall quips about this newbie system on the block called UNIX, and its having roots in Multics.
[Dave Vinograd] X.25 was added to Multics at the request of Canada in support of Transpac. I did a deal with HIS Canada (Wally Halladay) to do X.25 (which I wanted to do anyway) for a number of VIP7800 terminals which CISL was desperate to have.
[Randy Oyagi] I will still try to find some info on our L68 6 CPU x 4 SCU x 2 IOM system, resplendent with 451s (of course I kept the address plug of the RPV) and 501s. I have in effect stayed with the same company. Bell Canada swapped some personnel with IBM (me included), and I now work for the vendor that I fought for so many years -- you know, trying to get management to understand the importance and superiority of Multics. Of course, back then, batch was just about the only thing that counted, and if your time-sharing wasn't full-screen it wasn't any good. The biggest insult, to me, is when you try to explain that you worked on Multics, the mother of UNIX, no one cares. But the whole world has to stop to learn that Madonna begat Lourdes.
[John Potts] I was on site at Bell-T for over two years, Level 68 processors serial 20 and 21. The back panel was so bad they both were sent back to Phoenix for a rewrap and stage...... The original L68 went to the NEC processor, I forget the dates.
[Joe Caruso] I had three roles - System Developer plus Benchmarks + Capacity Planner; I had a nifty algorithm that was pretty accurate for planning how many users we could run with x processors and y MB of memory. I seem to remember that an Ernie Reimer was the first HIS FE onsite. The configuration in Don Mills went to 6CPUs and 12MB! of memory. We also had visits from a number of MIT types to help us tune the system; John Bongiovanni was one (Bongo Code - the scheduler I believe); Bob Mullen helped us with tuning - drank cases of Pepsi while on the computer floor and tutored us on the innards of Multics over 2-3 days.
Happy Happy joy joy!!
[Joe Caruso] I broke my leg skiing in Vermont after one of the Multics training courses in Montreal, and missed out on a the next few. I took the last two in Phoenix in the summer after I got off my crutches. I stayed at The Pointe with most of the other students. We worked diligently on our assignments at the end of the day and then hurried back to The Pointe to make happy hour. We would get 'happy' on the free margaritas, and about 8pm head off for dinner.
Heart stopping moments
[Joe Caruso] The doorway and subsequent pathway onto the computer floor in Don Mills was behind a row of our Multics Disk Drives. It was the same path used by the IBM operators to bring in skids of paper for the high speed printers. While visiting the floor, I often winced as they struggled to bring the skids on the power dollys past those drives just dreading that they might hit one of the drives.
Clean Power? We don't need no stinking clean power!
[Joe Caruso] Our Don Mills processors had some very weird hardware problems shortly after they were installed - no rhyme or reason. The Bell janitor quietly brought it to our attention that the system was not running on clean power from the UPS - it was on raw power. A formal power supply and wiring audit was ordered and the problem was formally documented and repaired.
Everything old is new again
[Joe Caruso] I get a kick out of showing my son, who's a PC geek, features of PC sw and hw technology that were in Multics. The 486 processors had a ring structure very reminiscent of Multics; in XP, the idle process which runs when there's no work to do just brings back old memories.
I can do anything I chose to
[Joe Caruso] In the early days of Multics, Bell established a policy of no CPU charges after 5pm, to entice people to run batch jobs off hours, maybe work after hours to reduce the load during daytime. Well, one enterprising fellow worked after 5pm to develop a rather large application which required an additional CPU. Well, the Multics Steering Committee got in a snit, chastising him and telling him he was not allowed to do this. His cool response was, but I have - how can you sit there and tell me I can't? We added the additional CPU.
Who needs a map to Billerica?
[Joe Caruso] There was a session in Billerica to discuss some X.25, I attended along with Kamil Khan, Stan Drake, Doug Spencer, Doug Cooke, Ron Brown, and another HIS Toronto exec, who was chauffeuring Kamil from Boston to Billerica. Kamil and driver got lost and showed up an hour late. Yes, people were getting nervous because it had rained that morning and we did see a few cars sliding into each other on the expressway didn't we Doug? One of the Bell Execs got a surprise at the hotel the night before - he walked in to his room and saw a lovely red evening dress laid out on the bed, "Discretion being the better part of Valour..." he said, he returned to the front desk to get a different room.
A Trip to MIT: 1982?
[Joe Caruso] Northern Telecom had developed a training package for their SP1 telco switches on the Multics Platform. Nortel was eager to improve their sales in the US and saw the installation of the training package on a US based system as a bonus. I was sent to MIT to do the install. We did hit one problem - there was a bug in the Terminal Configuration file that incorrectly processed an attribute; I remember the CISL developer reading the code on the printer as it was coming out, he stopped the printer before the listing was finished printing, ripped off the listing, went away, and had it fixed within 30 minutes. The training software was successfully installed within 3 days. The real highlight of the trip was seeing an arcade style game on running on an IBM PC.
[Joe Caruso] I still have a red Multics T-shirt, Frisbee, pin, and sticky labels. I also have the quick reference book Commands and Active functions.
Multics - Many Unused Large Tables in Core Simultaneously
[Peter Langenberg] This is of course what we referred to Multics as we were testing and benchmarking it for replacement of the GCOS Timesharing System. Though the system was not much faster than the basic GCOS system, we would be able to expand it and that was one of the key reasons that we wanted it so badly. This was the beginning of personal computing, long before the PC XT and AT that came along shortly after.
[Peter Langenberg] Interesting story for the days when we would be able to go for a week to a far away city at the company's expense and actually take in technical sessions. Had the opportunity to attend a session in Detroit with a few of the Operations folks and one was Bill Harris. We had all had a few beer first and then decided to have coffee and desert before heading to bed. Bill asked for "hot apple pie and ice cream". The waiter indicated that they did not have that in the restaurant. This was a Delta Hotel and Bill quite calmly asked the waiter if they had "ice cream" - answer Yes; then asked if they had "apple pie" answer was Yes; and then he instructed the waiter to find a microwave in one of the 500 rooms and warm the apple pie and drop the icecream on top of it and get it back here really quickly.
Bell-T, Don Mills: 1990
Bell-M, Dorval: 1993Information from Wilf Mandel, Julie Jackson Sibert, Dave Vinograd, Randy Oyagi, John Potts, Joe Caruso, Alan Sawyer Peter Langenberg.