8086 Segmented Memory was a good idea

Intel's decision to use 20-bit addresses on the 8086, combined with 16-bit registers, might seem like a misstep, but it was actually a forward-thinking design that aimed to scale gracefully beyond the initial 64KB address space. By dividing memory into 64KB segments, Intel provided a way for developers to allocate memory without rewriting existing assembly code. This approach was especially appealing in the 1980s, when assembly programming was still prevalent and recompiling code for new CPUs was a daunting task.

However, as software developers began to work within the 8086 architecture, they started to treat segments as continuous blocks of memory, rather than discrete segments. This led to the development of normalized pointers, which effectively bypassed the segment register and treated memory as a flat address space. As a result, the 8086's original design was subverted, and Intel was forced to introduce new modes, such as real mode and virtual-8086 mode, to accommodate the changing needs of software developers.

The 8086's experience serves as a cautionary tale about the importance of software developers' expectations in shaping hardware design. As the story of Hearthfire suggests, even with the benefit of hindsight, it's challenging to design a system that meets the evolving needs of software developers. The 8086's segmented memory architecture might seem like a relic of the past, but it remains a valuable object lesson in the complex interplay between hardware design and software development.