On 17/05/2018 10:33, Bob Eager wrote:
On Thu, 17 May 2018 09:02:19 +0100, Brian Gaff wrote:

Well, is not ICL now part of Fujitsu?
There are still a lot of old mainframes etc floating about in museums,
though whether any work or can be used to demonstrate anything is
debatable.
Brian

There's a working Elliott 803 at NMoC, and they have a 2966 almost
working - won't be long now.


The LOndon Hospital computer centre had an Elliot 803. Apparently
it had tape drives where the tape was like a 35 mm film, with sprocket
holes.

That was then replaced by a Univac 418-III in the late 60's, with a
massive horizontal drum called a FastRand, where the read write heads
were a long bar that was moved from side to side across the drum,
which was about 6 feet wide. It also had 4 'spin dryer' drums that
allowed an access time of 4.25 millseconds which was fast for that
era. These could never be powered off because they developed a
vibration at a particular speed which wrecked the bearings. The card
reader was the size of car. Needed a 400 Mhz power supply which occupied
a room all on its own.

It was an 18 bit one's complement machine, so Octal 777776 was -1 and
777777 was minus zero which generated an error stop. Also, no stack.
Calling a subroutine meant using an assembler instruction SLJ (sludge)
- store location and jump, so you could go to any depth you liked,
except that it only had 128 *K* words of 18-bit memory so the London
Hospital real-time system was an RTOS batch job that ran all day and
within that the 16 available files were logically mapped to provide
multiple inhouse database files, with transaction processing and full
before and after journalling and logical transactions. 128 Uniscope
VDUs and printers around the hospital provided 24/7 services to all
the wards and clinics in the 1970's when few commercial programmers
had even heard of the concept of a logical transaction.

All written in Univac assembler and developed using punched cards by
hospital staff programmers. :-)