I'm finding D. F. Parkhill's 1966 book "The Challenge of the Computer Utility"[0] to be endlessly fascinating. It provides quite a detailed snapshot of the state of computer development. It was written at the tail end of the "Patent wars of 1962–1966"[1], yet there's no mention of integrated circuits, nor "Micrologic" (an early name for ICs).

In the discussion on memory technology, the book contains several tables as two-page spreads, which are awkward to assimilate when reading the book as a PDF on a small screen. I've included the relevant pages here as two-up images for my own benefit as much as yours!

The information summarised in the tables is fascinating. In the table "Classification of Memories by Function", much of the language is familiar to a modern reader (compared to, say 1940s discussions of 'organs'); CPU registers, RAM and persistent storage are recognisable:

Storage register: Usually a one- or two-word memory used for the temporary storage of some quantity before it is transferred to another memory or circuit; i.e., accumulator register, multiplicand register, index register, etc.

Internal working memory: The main working memory of the computer, in which intermediate results and instructions are stored.

Mass data memory: A high-capacity storage system, external to but under the control of the computer, used for the storage of bulk data such as tables, files, and sub-routines.

...yet the "Classification of Memories by Operating Characteristics" reflect mid-1960s (or older) technologies:

Regenerative: A memory whose contents gradually vanish unless they are periodically regenerated, e.g., a Williams tube.

Modern RAM is of this type!

Read only: A memory whose contents can be changed, if at all, only by off-line human intervention, usually involving rewiring, the removal or insertion of plugs or the punching of holes, e.g., a card capacitor store, diode matrix, etc.

"Memory Devices" is fascinating. Here's just one row:

Type: Magnetic core
Physical Principle: remanent magnetization on small cores of square hysteresis-loop
ferrite material
Application: high-speed internal memory, registers, and buffers
Status: standard memory for majority of all computers in all price classes
Remarks: in addition to the normal coincident-current destructive readout single-core/bit systems embodiments there are also multiaperture core systems such as Biax and the transfluxor systems, and multiple-core/bit systems also wired-core read-only systems

Biax? Transfluxor systems?

...and what on earth was "Magnetic rod memory"?

Type: Magnetic rod
Physical Principle: magnetic coupling via removable ferrite rods between loops in a
woven mesh
Application: read-only very high-speed auxiliary internal store
Status: in use on Univ. of Manchester MUSE, Ferranti ATLAS computers, and several Italian machines
Remarks: retains advantages of wired-core memories but permits easy modification. Highest speed operating memory of comparable size to date, (0.15 microsecond access time, 8192 words, 48 bits)

...sounds promising! Why haven't I heard of this technology? Check Wikipedia[2]:

Rod memory is one of the many variations on magnetic core memory that attempts to lower costs by automating its manufacturing. [...] Like many similar concepts [...] rod memory was competing for the role of taking over from core when the first semiconductor memory systems wiped out the entire market in 1970.

Oof.

I'm finding it very much worthwhile to read not only older histories of computing, but also old books that provide a survey of the state of the art at a given time.

[0] https://archive.org/details/challengeofcompu0000park
[1] https://en.wikipedia.org/wiki/Invention_of_the_integrated_circuit#Patent_wars_of_1962–1966
[2] https://en.m.wikipedia.org/wiki/Rod_memory