TECH Signal 527 2 feeds carried it
Flat and segmented memory models reveal recursive subdivision in software
↗
Illustration only Photo by Robin Glauser on Unsplash
Engineers see that memory subdivision is inherently recursive, challenging hardware that prefers fixed-depth structures.
The hardware-software mismatch means processors must either bound decomposition depth or rely on software to enforce finer bounds. Adopting models like CHERI shifts the burden to software, requiring explicit capability passing and potentially breaking existing code. Widespread use depends on the industry accepting the externalized development effort needed for hardware to gain ubiquity.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Programmers naturally apply recursive subdivision to flat memory through arenas, pools, and nested struct fields.
Traditional x86 segmentation offers only coarse-grained rings, limiting protection to user versus system rather than fine-grained confinement.
CHERI provides fine-grained, recursively applicable bounds via capabilities that software narrows and propagates, avoiding hardware depth limits.
THE READ
What the cluster adds up to.
The article observes that when programmers work with a flat address space they instinctively create recursive subdivisions such as memory arenas, pools, and nested structure fields. This reflects a deeper habit of breaking memory into smaller and smaller regions to manage allocation and lifetime. Hardware, however, is described as conceptually finite-state and tends to favor fixed structures with bounded depth. Thus a mismatch appears between the recursive nature of software memory organization and the non-recursive nature of typical processors.
Traditional x86 segmentation was examined as an early attempt to provide non-flat memory but is shown to give only coarse-grained protection. The four-ring privilege model separates user from system but does not prevent a segment from being reloaded to reach any other segment. Consequently, segmentation isolates faults rather than enforcing fine-grained security against malicious code. The author notes that this limits the usefulness of segmentation for modern confinement needs.
CHERI is presented as a solution that retains hardware flexibility while allowing software to handle recursion. Capabilities carry bounds that can be narrowed arbitrarily by software and passed along, with the requirement that each derivation step reduces the bound monotonically. This means the hardware does not need to impose a fixed depth limit; the effective reachable memory emerges from the chain of capabilities. The approach therefore aligns hardware with the recursive subdivision pattern used in programs.
Adopting CHERI or similar mechanisms faces friction because existing codebases expect the freedom to perform non-monotonic address calculations. Overcoming this requires development effort that, if the hardware gains widespread use, becomes an externalized cost across the industry. The payoff is stronger safety and security guarantees, as hardware can enforce fine-grained confinement without sacrificing flexibility. Whether the industry accepts this trade-off will determine if recursive memory models become mainstream.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER