Leaving IRCAM in 1984

"Date: September 21st 1984
Duties: Computer Services Manager
Reason for leaving: Looking for a more research-oriented position

Adrian Freed is an extremely competent - and recognized - UNIX expert. Being always innovative and well organized, he was a very efficient group leader.

Adrian Freed was always available when needed, including off-hour moments (late at night or at weekends) when difficult tasks arose"

Jean-Pierre Keller"

The scientific director of IRCAM sent the above brief reference to Bell Labs, my employer after I was Responsable du Service Informatique at IRCAM. I found it in my archives while writing a CV. This reference is strikingly different to the implication about my departure from IRCAM reported (on page 254) in Georgina Born's pioneering book Rationalizing Culture where I am coded in the text as FA.

I have far more confidence in Georgina's recording of the events and in the various supporting documents than my own memory so it has been a fun puzzle to try to untangle from the record these different interpretations of my time at IRCAM. One thing emerges very quickly: Georgina's accuracy with respect to what she observed is impressive. She is a great ethnographer. A challenge was that she couldn't be at every meeting or interview everyone on an issue so any problems with her account are likely from errors of omission. She did go to meetings of the artistic committee where many grumbles about the availability of computing resources to complete pieces were voiced. She didn't attend scientific committee meetings so how these grumbles were discussed or responded to is not described very much in the book.

It might be interesting for someone to dig into the IRCAM archives and look at the minutes of those meetings. I will attest to their accuracy as I was responsible with Margaret Tunstill in writing them as secretary of that committee. This was something Pierre Boulez asked me to do because he liked my confident and responsible challenges to the 4X group who were overpromising and underdelivering. He had trouble finding an accepted or acceptable scientific director during this period to manage these issues so the committee was a rather clumsy apparatus to try to keep things honest. This pattern of overpromising and underdelivering was rather ubiquitous and obvious to me the first day on the job. I was hired to replace their PDP10 with a Vax 11/780 they had ordered. On arrival I reviewed the order and concluded their assumptions about what the Vax 11/780 and a new and unfamiliar and untested operating system UNIX BSD 4.1 could do were naive and optimistic. The system they ordered was configured so minimally it couldn't even run the operating system well - too little memory, too little I/O and inadequate disk space. I addressed this by creating a realistic budget, ordering the required extra parts and extending the operations of the PDP10 by the many months it took to bring the Vax 11/780 into operation so that pieces in process could finish their production.

Georgina does a good job of describing some of the resource contention issues at play in this period.

I will take these up in a future note because it is interesting that many of the core disagreements are still active in 2019.

It took a while for me to realize that the management structures in place were not going to produce the conditions where the Service Informatique could do its job well. One event that confirmed this for me and outlined my limited agency was when the programmer in my group, the late, great Patrick Sinz, told me he had been reassigned to the 4X group. This produced the need to bring the American consultants in to do the system software upgrades that Georgina refers to. I still don't know how that decision was made without consulting me but usually these sorts of things came about from private conversations with Pierre Boulez outside of the committees. It would have likely been Jean-Pierre Keller, David Wessel and Jean-Pierre Armand who brokered this particular compromise. This is an indication that points to the classical difference between what happens in practice and what diagrams like an organogram might suggest should happen.

The bar for how well a Service Informatique could run was set rather high for me by a mentor in college, Ian Johnstone, who hired me while I was a student into his group supporting computing for the Australian Graduate School of Management. Because of better funding and his generous hiring they had the best computing environment on the UNSW campus. Ian was an outstanding system's programmer and the PDP 11/70 he ran served an unusually large number of concurrent users with a very diverse range of software. I remember being disappointed with how poorly the Vax 11/780 at IRCAM ran compared to the supposedly inferior and much cheaper computer, the PDP 11/70. It actually took two major upgrades of BSD UNIX 4.3 before the Vax was really comparable. Many of Ian's performance innovations were integrated into those kernels. I mention this as it points to another omission in Georgina's account which emphasizes the reliance on US hardware and software. I was actually trained in Australia and many of the innovations introduced into IRCAM's software base can be traced there including the graphics package a college buddy installed. Also Patrick Sinz and the other french "squatters" were responsible for important software too - notably the integration of networking features as IRCAM pioneered participation in the emerging research internet and TRANSPAC.

I will close by outlining some facts about my departure:

My last month at IRCAM was July of 1984. I had to give 6 months written notice to leave at this time so the connection Georgina implies between an "avertisement" (warning) and leaving "two months" later is false.

I was involved in finding my replacement, Robert Gross, who sensibly made a condition of his employment that a second machine be purchased to stage system improvements on (a Vax 11/750). His eventual replacement required further equipment purchases. It took 4 years worth of budget to bring the computing resources to the point that research and production activities at IRCAM could be said to meet the promises and expectations of the house.

I shared my disgruntlement with the working conditions at IRCAM with my mentor Ian Johnstone who had moved from Australia to Bell Labs. He recommended me to a HCI research group at Bell Labs. who offered me a job - with a promise I could do research. The disruption in my team, this offer and an incident when I was sexually harassed at IRCAM are why I gave my notice and left.