Computer history: data communication (1974 – 1978)

Data communication and interactive work: the first beginning

Around 1975, the data communication facilities of the Physics Laboratory TNO were very limited. The data communication controller could handle a maximum of eighth synchronous and/or asynchronous lines, provided that the total baud rate did not exceed 30-40 kbps. Above that rate, the very complex peripheral processor program (1MR) could not keep up with the data streams. Both the Laboratory for Electronic Developments of the Armed Forces (LEOK), the TNO Institute for Sensory Physiology (IZF) in Soesterberg, and the Prins Maurits Laboratory (PML) in Rijswijk were connected to the Physics Laboratory (PhL) with synchronous 2400 baud and 4800 baud modems. The very expensive, many kilos heavy modems were full of coils and knobs to tune the interconnection characteristics. Every month, a PTT employee tuned the coils.

A modified version of the UT-200 (UT200) (Mode 4) synchronous protocol was run on the Digital Equipment PDP systems (PDP = Programmed Data Processor) at the said institutes so that these institutes could submit batch jobs and receive print output. This can be seen as the rudimentary start of the network between the TNO Defence Research Institutes, a network that was called later the Defence Research Network (DOnet).

Working interactively: synchronous hick-ups

In the specially designed terminal space of the PhL there were six synchronous CDC 711-terminals. Synchronous meant that the terminal was polled every few seconds to check whether a ‘send screen’ command had been given. Output towards the terminal occurred block wise. These 711 terminals were delicate devices of about 15 kilos each. Two to three terminals were connected per synchronous line of the data communication controller. A neighbour who received a lot of output to his terminal meant that the response time of all terminals went down. The user could reserve a terminal for a maximum of 15 minutes. At the end of the quarter, one was allowed to reserve a new time slot.
In order not to be entirely dependent on the large printer and card reader, the Laboratory also purchased a CDC 734 batch station consisting of a console, small card reader and printer. This CDC 734 also worked on the basis of the UT-200 synchronous protocol.

A network overview of the Physical Laboratory from April 1978 shows the coupling of the Calma (“ECAD” system), seven synchronous terminals (711s) via two lines, the PDP 11/45 of the PML (with the MUX200 package), a 711 terminal at the Higher Military School (HKS), the 734 batch station – all with 4800 baud synchronous lines operating via the UT-200 protocol – and the PDP at the LEOK via a 2400 baud synchronous connection. In addition, 8 Newbury terminals (300 baud), one Tektronix Silent 700 ‘movable’ terminal (12 kilos) with the built-in telephone modem and a Tektronix 4010 graphics terminal (1200 baud) were connected via the asynchronous lines. Later that year a Tektronix 4014 graphics terminal was connected for displaying radar coverage images using the PLOT-10 package.

SUEDI: Single User Editor IWIS

The standard Control Data file editor was a multi-user, multi-threaded editor that occupied roughly 40000B words (123 KB) of memory. This was a burden for the system! The TNO Mathematics and Statistics Institute (IWIS-TNO) had a Control Data system as well. They developed the “Single User Editor IWIS-TNO” (SUEDI). This editor was written in a mix of Fortran and assembler (COMPASS) code. This editor was faster, had more features, used far less memory and -above all- when a crash occurred only the single user was affected.

Talking about malfunctions: interactive work required reservation of a synchronous terminal for 15 minutes, hope for a fair response by the system and a synchronous 711 terminal that kept alive as well as no hangs of the operating system. A good habit was to save all changes each five to ten minutes to keep the losses low. And Murphy was always around … During a couple of years, systems programming worked a lot on improving SUEDI by adding new features, improving the speed and reducing the need for memory. The latter required the split of program subroutines in even smaller modules that could be loaded as separate overlays. The largest overlay was the one who’s size had to be decreased down to the next lower multiple of 100B words. It required sometimes multiple days to figure out a way to decrease the largest overlay by two words (4-8 hardware instructions). Often, a solution turned out to be a gain of 100-200B! At that time, of course, one had to look at the next largest overlay with a fresh look… In a couple of years, the SUEDI program was halved to 2700B words (11 KB). Colleagues at the computer centre of the Rijksuniversiteit Groningen implemented new search methods for text fragments, such as the Boyer-Moore algorithm, which resulted in an acceleration by a factor of two to ten for searching for a text fragment, depending on the text fragment.

Around 1980, in parallel, an (almost) complete ASCII-version odf SUEDI was developed by IWIS-TNO based on the old code base. Two 6-bit bytes were used for the storage of one ASCII character, allowing texts with small letters and special characters. Later on, Daan Sandee developed a fast and very small joint code base for both versions of this editor: SUEDI and SUEDA. Because of their rich feature set, their fastness and the small memory occupation, the SUEDI and SUEDA programs became very popular at almost all Netherlands computer centres that used CDC-systems. New SUEDI/SUEDA-developments, system corrections and new features were exchanged during INFOSYS-meeting (Information exchange Systems programming) that were held bi-monthly. In this manner, sites helped each other. Interesting system and application codes were quickly adopted and then often developed further. One could actually regard part of the meetings as the first ‘shareware’ sessions.

The by CERN, Switzerland developed SCAN-program was converted by the NLR to the NOS/BE-platform. The earlier mentioned Booyers-Moore algorithm was added to it by the Rijksuniversiteit Groningen and the Physics Laboratory RVO-TNO added the full-ASCII feature.

For example, the SCAN program acquired from CERN, Switzerland was converted by the NLR to the NOS/BE platform, the previously mentioned Boyers-Moore algorithm was introduced into the SCAN program by the University of Groningen, and the Physical Laboratory RVO-TNO made the program suitable for the complete ASCII set. The former IWIS-TNO provided the Tekstopmaker IWIS (TOI). TOI created neatly formatted output based on what we now call ‘HTML-like’ formatting instructions.

In addition to the INFOSYS meetings, the heads of Operations met as well on a regular basis. They discussed problems with suppliers of computer media, printer paper, operational problems, emergency support in case of a system break-down. Data exchange sometimes gave problems as well. Jointly, they developed courses for new operators. These meetings were hosted on a round-robin basis by the various Netherlands sites using Control Data equipment (at that time): Control Data equipment (ACB, ACCU (ACademisch rekenCentrum Utrecht), Control Data Nederland, Energiecentrum Nederland Rekencentrum (ENR), TNO IWIS, KEMA, Lips, MID, NIKHEF, Nationaal Lucht- en Ruimtevaartcentrum (NLR), Rekencentrum Rijksuniversiteit Groningen (RUG), Stichting Academisch Rekencentrum Amsterdam (SARA), NATO’s Shape Technical Centre (STC), the Ministerie van VROM and the Physics Laboratory TNO.