The unknown hackers (2000)
salon.com
salon.com
They probably refer to a 386+ way to manage memory - pagination - that allows to do much better.
The OS considers memory as a number of (originally 4kB) pages, some occupied, some others free. When you need a larger chunk of free memory, the OS organizes a set of pages to assemble a segment of memory. The application sees a regular linear memory, but the OS is not left with pieces of unrelated small memory spaces: the OS keeps a directory of free pages that can be used in any order to create the next "any size" linear block.
When the free page stock is exhausted, the OS keeps statistics of the least critical pages and unloads them to disk (OS code for unload destination, but it is the processor that provides the triggers), generating free memory pages.
The ghost pages are marked as occupied in the segment description, but unavailable (no RAM behind). When a program calls such a page, a processor exception calls an OS routine that reloads the missing page to a free page and connects it to the "linear" memory segment, as seen by the program. Then program execution is resumed.
The identification of the least critical pages is a major element in OS performance, since the unload/reload mechanism makes memory access work at hard disk speed :-( [Off-topic: pagination/segmentation are great, but I am more caution about the swapping mechanism, as it has made programmers very lazy in memory management, and previous computer generation very slow, mostly during the 1990s and 2000s - when memory was so expensive.]
Keywords if you like to know more: LDT, GDT, Descriptor Table, segmentation, pagination...
This is a historical curiosity from PDP-11 before the advent of paging, but apparently still used by some malloc() implementations. Modern memory management doesn't use linear mapping but instead uses paging to create chunks of accessible memory. Some applications still use a big linear address space by appropriately asking the OS then interoperating with memory management to only request the memory that is needed
Aaaand the tab is closed as my expectations for the story were some hyperbole nonsense.
"An Article about Bill and Lynne Jolitz who wrote an open-source Unix pretty much at the same time Linus Torvalds did but failed to gain any traction" for anyone else who still has an open mind for reading. ;)
Among not too young users of FreeBSD or other BSD derivatives, it is well known that those operating systems have been partially based on 386BSD, which was written by the Jolitzes. Their role for BSDs is indeed as important as the role of Linus for Linux.
While even inside the BSD community they are not as well known as Linus is elsewhere, that is normal, because their contributions ended after the first years, while Linus remained very active until now.
I have started to use both FreeBSD and Linux in 1995, and until around 2003/2004 FreeBSD was much better.
Then the BSD derivatives had very slow transitions to being able to handle multi-threaded and multi-core processors and also to handle more modern peripherals, e.g. GPUs, while Linux improved much faster than them.
And I am certain that there are "more famous" (whatever that might mean) programmers individuals do not know because of different interests and disciplines.