Project Flower was started in 1984 in response to Red Phillips' concern that there was a lack of real world input to the corporate laboratories. Red had been a hardware engineer in the 1950s and had risen through the ranks at UNIVAC to become head of Product Development. In the early 80s, as a result of organizational conflicts in what was then Sperry Univac, he joined Honeywell as a staff VP reporting to Jerry Meyer (who was HIS Systems Executive VP, or some such title, and a former UNIVAC executive as well). Red felt that the chip development at the Semiconductor Division, the packaging development in the Physical Sciences Center and the logic design synthesis work in the Computer Sciences Center all needed some kind of product focused technology driver.
I had worked in Red's organization at UNIVAC and was on-loan to him at HIS headquarters in Minneapolis. He and I held several "brain-storming" sessions with managers from the laboratories and they agreed to sponsor an "advanced development project" focused on creating a Multics follow-on system. Since we were to be structured as a "skunk works," we chose the name "Flower" after the skunk in the movie Bambi.
The team members were (my apologies for bad spelling or forgetfulness):
- CISL (Boston) - Benson Margulies
- Multics (Phoenix) - Rich Coppola, Jim Dahl, Dick Demers, Ron Lange
- SSED - Jim White, John Winkler
- CCSC (Minneapolis) - Rich Cloutier
- CPSC (Minneapolis) - Ron(?) Cummins(?)
- HIS staff (Minneapolis) - Tom Rykken
It was my first attempt at trying to run a distributed project -- however, Multics "forum" turned out to be an ideal project management tool, providing both a means of constant communication and rapid issue resolution, as well as creating an on-line "project notebook." Phoenix engineers were very early adopters of the Apple Macintosh and we used Beta copies of the software for these machines for drawings and presentation materials. Unfortunately, I can't find any archives of this material.
The system design approach was to use the memory and I/O components of RPM that were under development at that time (CR66 I/O was the I/O subsystem).
It took a couple of months to get a firm architecture definition (we threw out anything in the 6000 architecture that Multics didn't need and added in a few desired features). In parallel, we completed the design for the CPU instruction-sequencing pipeline. After a few more months it looked like we could actually complete an implementation and I received a promise of funding from JJ Renier(Pres/CEO).
Somewhere during this time the Multics group was put directly under Jerry Meyer. I understood that this was intended as a means of "protecting" the group while Renier and his staff decided where to put it for the long term. However, I suspect some of the Multics senior management interpreted it differently.
In any case, the project was then put under the Office Systems Division in Boston that was being managed by Gene Manno. At that time, Jim Hannigan, who was in the System Design group in Boston, took over the non-technical project management role and began building a promising business case for the product. We continued to make good progress to the point that we were beginning to worry about getting on the chip personalization schedule at SSED.
However, the "subjugation" of Multics under Manno did not go well. When a budget crunch hit, project Flower lacked sufficient political backing in any quarter and it was killed. I wasn't surprised, but felt betrayed and soon left HIS to go back to Sperry Univac. In a year, the merger with Burroughs that created Unisys led to my return to what became Honeywell-Bull.
Posted to feb_wwide 08 Feb 2001