Back Up And Describe It Again
The JFE Bernoulli Disk item in 40 Years Ago (Jan. 24, p. 54) was interesting. But it wasn't a "floppy disk," although the disk material is similar. Note the rotational speed, "up to 15,000 RPM," faster than most hard drives. In floppy disks, the head slides directly on the disk, so they must use a very slow rotational speed to avoid destroying the magnetic coating. Bernoulli disk drives use aerodynamics to pull a rapidly rotating flexible disk close to, but not touching, the head. A dust particle would distort the airflow, and the disk would pull away from the head. So head crashes and scratched disks were much less likely than with hard drives.
Bernoulli drives have generally been used with removable disks as backup devices, and positioned between hard and floppy drives in speed, capacity, and price. I believe Iomega sold them for years, but I don't know if their ZIP drives are a new implementation of the same principle or something else.
It was surprising to see a Bernoulli disk from 1960. I first heard of them nearly 20 years later. The picture is apparently of a multiple-disk unit. The description doesn't mention the head moving. Since it appears to have been intended to compete with drums, it probably wasn't designed for easy cartridge removal, and may have had just one head per track and one track per disk. Drums were about to lose their market to core memory (faster) and hard drives (much higher capacity), so in hindsight, this was the worst possible market positioning.
Keep your eyes peeled for the 40 Years Ago coming in the April 17 issue. It will follow up this Bernoulli item.—Ed.
C++ Isn't Really Where It's At
I must applaud you and Gary Smith for pointing out the pitfalls of C++ ("Languages Fight To Take Over As The Next Design Solution," Jan. 24, p. 68). Mr. Smith finally told the truth to the engineering community. It seems that C++ actually was a failure. That's why the standard language today is C/C++. C++ adds a lot to C. But try to use it as a standalone object-oriented design language and it doesn't work too well.
Quality is a major concern. Unfortunately, no mention was made of the one language that really meets your requirements, Ada. Ada is an ISO standard (ISO/IEC 8652:1995) that has extensive concurrency constructs including: rate-monotonic scheduling, the ability to treat bit patterns as standard data types, a superlative object model, and an unparalleled record of reliability.
Ada and VHDL were initially supposed to be the same language. In fact, M. Mills and G. Peterson have demonstrated that a mapping of VHDL to Ada is reasonably straightforward in their paper, "Hardware/Software Co-design: VHDL and Ada 95 Code Migration and Integrated Analysis," Proceedings ACM SIGAda Annual International Conference 1998, ACM Press, p. 18-27.
The parts of VHDL that are not in Ada could be added as annexes to the Ada standard. In fact, since translators can be created between VHDL and Ada, the two languages should be merged.
And The Credit Goes To...
Your article brought to mind a saying someone once told me ("Revel In Your New Ideas Despite Negative Feedback," Feb. 7, p. 164):
"You can accomplish just about anything in life if you don't mind who gets the credit."