The Jim Austin Computer Collection

ICL System 25

This machine was donated by Mark Norman in 1998. The machine was used with 12 or so terminals to provide Sales ledger, Bought ledger, inventory, and some purchasing functions for Tiffany Sharwoods up to about 1992.

The following has been provided by Brendon Murray about this machine, Feb 2008.

I came across your web site & I notice you have very little information on the ICL System 25. I worked on the prototype of this machine, shortly before it was released, verifying that the packages from its predecessor, the ICL System Ten (acquired from Singer) would run correctly. This work was done in Feltham, in 1980-81. I think it was released in early 1981, but it may have been late 1980. I moved to Denmark shortly after this project, to work for an ICL dealer, Kontor Automation A/S (later became F.E.Bording A/S), where I looked after the system side of the System 25, as well as any unusual projects. I actually still have a "Programmers Reference Booklet" for the System 25 - I have toyed with the idea of creating an emulator at different times, but always found something useful to do with my time instead.

The machine's name is based on the instruction set. Originally it was called System 29, but there were some issues with the microcode so four of the instructions were dropped, and the resulting 25 became the name. Later they resolved these issues and produced an "extended facility processor" for which they got all 29 instructions working and even added a further six, a total of 35 instructions. I think they renamed this machine the "System 25 Plus".

Early models of the machine had front-loading plastic-encapsulated removable hard disks. However there was a flaw in the design (which was later rectified) where powering off the system caused a spiral write to be executed on the disk drive, rendering the thing unreadable.

The System 25 was a decimal-based machine, where each instruction was 10 bytes, aligned on a decimal boundary. Mathematical operations worked on decimal data rather than their binary equivalents. Memory was partitioned, where a task would typically use a partition (max size = 80KB, although 2KB was reserved for system use), with up to twenty partitions in a system. Shared memory was in "Common", which could be up to 320KB. Of course, the OS was loaded into Common, being booted from partition zero by typing "0001" at the command line. The OS was located in the lower 80KB of common - I can't remember how large it was - and the rest of common was used to optimise the partition-based programs, where common functions were loaded and called from the partitions. The character set used was ASCII, i.e. ISO-646 with the usual national variants for Europe - I don't ever remember an Asian or BiDi version being produced.

The OS was called DMF III (Disk Management Facility?) and used loadable modules to access the disk called Logical Input-Output Control System (LIOCS). The system commands typed at the screen were referred to as CSM (the M stood for Manager, but I can't remember what the C and S were).

Programming the System 25, like its predecessor the System Ten, was pretty close to the metal. While ICL did produce a COBOL compiler later, along with an enormously complex framework called IAS, the System 25 remained popular because of its simplicity. Any programmer worth his/her salt knew exactly what the individual bits in each 10-byte  instruction did, and could assemble an instruction by hand (and often in his/her head). However, there was an assembler program called ASMIII (assembler 3) and I vaguely remember an independent assembler called ASMIV, but I think that sank without trace.

While ICL in the UK pushed this machine as a retail controller, it was actually used in a very broad range of applications. Those I had direct contact with were:
 

• A single S25 controlling 150 ICL 9500 POS retail terminals in Illum, Copenhagen (at that time a subsidiary of the UK House of Frazer), and even drove a card-embossing terminal. It also processed transactions and uploaded them to the mainframe, and uploaded credit-card transactions to the financial institutions (as well as credit checks on the cards).
• Arbitrage desk where 10 System 25 screens switched between a number of financial information services, i.e. it was primarily a comms controller, connecting to a number of services using different protocols. I remember that one service had set up the connection on the IBM mainframe incorrectly: when we realised the problem, the guy at the other end apologised and explained that it would be a week before it could be corrected, so we simply reconfigured on the S25 side & were connected in less than a minute. The IBM guy was stunned that it was possible to do that so quickly, but that was really a measure of the machine's flexibility and simplicity.
• Many, many SMB installations where it handled traditional sales, orders, etc.
• Many factory installations where it was used in a production environment for resource planning, etc.

The machine itself was hugely flexible, including a very varied connectivity - this was very unusual at the time, where most small machines could only handle a very limited number of protocols and usually only had a single connection to the outside.

The following was contributed by Colin Davis in Oct 2009.

They were reliable machines, used for many years in Sainsburys. Some of the larger stores had the System 25 plus box. The configuration was 1 CSM terminal and 3 IAS terminals, connected via a patch panel. I cannot remember the spec for the box itself. It was great reading your web page. I supported these machines when employed by Sainsburys, theyhad one of these machines in all stores for managing their orders. CSM stands for Conversational System Manager.

The following was contributed by 'The Flying Dutchman'

>Early models of the machine had front-loading plastic-encapsulated removable hard disks. However there was a flaw in >the design (which was later rectified) where powering off the system caused a spiral write to be executed on the >disk drive, rendering the thing unreadable.

This problem was common with the System Ten 120 an entry level System Ten. This problem had us stumped for ages. When we eventually go the Modification instructions they suggested that we take an oil stone to the spindle while it was turning (instructions were given as to how to achieve this) at a precise angle. I whish I could get hold of a copy of these Modification Instructions. Grinding the edge of the cone was supposed to fix the problem. We couldn't believe that this would do it, the problem went away. I still can't believe it.

I worked in Spain from 1979 on. Though involved with 2900 systems we did have common peripherals so I had some contact with System 10 and System 25. Retail was not its strong point, Singer/ICl had happily got rid of Sears as a major customer. Retail would not take of again till the DRS20 systems became available when ICL won some of the major players such as Auchamp, L'Eclerc, Jumbo. When I arrived they were very strong in Time and Attendence and Production Control. This market was so important that terminals were manufactured in Spain to hold down the costs. Some large accounts were Pegaso-the lorry manufacturer, Ensidesa-the Spanish Steelworks. Well known Time and Attendence systems were the Ministry of Social Security and the Bank of Spain.

It was also used as a general purpose business system by many companies with one standing out in my memory; Philips Morris, the tabacco giant.