What Bench Instruments Owe to the First Public Terminals

One controller, several instruments, one printed log: a bench that measures by itself is a small network, and it runs on three ideas any shared machine needs. One piece of hardware is shared. One user has it at a time. A record says who did what, and in what order. Those ideas were worked out twice in the same decades, once at the bench and once by the people who put a machine in front of the public.

On the bench the arrangement is a bus. A cable runs from instrument to instrument, each unit answers to an address, and the controller speaks to one address at a time. A microwave bench fits the pattern closely: a source, a switch driver, a step attenuator and a power meter all hang on the same cable, and one program tells each of them what to do and reads the answer back.

The Local Loop, an independent journal on community networks and public internet access, keeps the public half of that record. Its guide to what was a free-net covers the systems that gave the public free dial-up access to local information and to the early internet, who built them, and the community pages they carried, and the journal follows the same thread into the library access points and the local builds of the present.

The bus that put a rack of instruments under one controller

Before the bus, an automated measurement was a stack of separate connections. Each instrument had its own interface and its own convention, and a rack wired that way behaved like a room full of people talking at once with no rule for taking turns. The bus replaced the stack with one connector and one protocol. It was developed at Hewlett-Packard in the 1960s, standardized in the middle of the 1970s as IEEE 488, and it is still known on the bench as GPIB or HP-IB.

The change it made was not speed. It was that a rack became addressable, and an addressable rack can be logged. A program can now step a source, set a switch state, read a power meter and write a row, all in one pass, and the row is the measurement.

Addressing is where a bench log is won or lost. Two instruments set to the same address produce a log that looks complete and is wrong, because the controller reads whichever unit answers and never learns that it asked the wrong one. The record shows a column of numbers with no gap in it. The same log also inherits a rule from the shared cable: one talker at a time. Instruments take turns, and the order is fixed by the program rather than negotiated by the units.

What was a free-net and how did people use one?

A free-net was a community computer the public could reach over an ordinary telephone line. A caller dialed in, was given a menu rather than a command prompt, and worked through local information, message boards. The constraints were exactly the constraints of a shared bench: one machine for many callers, a small number of lines, accounts instead of addresses, and a queue instead of a free-for-all. The journal documents what those systems were, who built them and which community pages they hosted.

Which was the first community computer network open to the public?

The record kept on that domain names the Cleveland Free-Net: a medical question board that grew into the first community network of its kind, and from there into tens of thousands of users. Its shape is the bench bus in another material: one machine, many users, an account standing in for an address, a rule for taking turns standing in for bus arbitration.

Did Houston have a free community network in the 1990s?

It did. The Greater Houston Free-Net offered residents free internet access from 1997 and hosted homepages for nonprofit organizations, which is how a charity or a parish could appear on the web before hosting was a product anyone sold. Behind the name stood a federation: the National Public Telecomputing Network licensed the Free-Net name and shipped the FreePort software, and the journal follows both its rise and its collapse.

What a bench keeps from that era

Three habits come out of the comparison, and all three are cheap. Address everything once, and write the map down where the next reader will find it. Expect the shared resource to be busy, and record the order in which it was used. Write the log for a reader who arrives later, not for the person who watched it being written.

On a microwave bench the habits look like a page taped inside the rack: which instrument sits at which address, which controller drives the switch, which meter reads the power, and the order of the steps. A calibration run that is not logged is a reading without a reference plane, and its result cannot be compared with the next session. What the run has to record is set out in insertion loss and isolation, and the vocabulary a log shares with a datasheet is in the glossary of control terms.

What a bus log holds, and what a missing entry costs later

Entry	What it fixes	What its absence costs
Address table	Which instrument answered which command	A complete looking column of numbers that belongs to another unit
Instrument list	The units present on the day of the run	A repeat measurement made on different hardware and read as the same
Step order	The sequence the controller used	A run that cannot be replayed, only approximated
Calibration state	Where the reference plane was at the time	Figures that cannot be compared with a later session
Date and operator	Who ran it and when	No way to date a drift, or to ask the person who was there

A rack wired this way measures the same component twice for the same reason a library kept its terminals in a row: the machine is expensive, the users are many, and an ordered record stands between the two. The assembly side of the same question, where the housing carries the control lines as well as the signal, is taken apart in what an integrated assembly contains.