Scalability, Attacks, Mobile

The paper's closing section looks past the theorem to three pressures on the networked world, and titles that future branded component systems — the same name this whole stack wears. Read each pressure as a stance. Scalability: growing a system needs coordination among the new resources, and coordination pays the very CAP bill the theorem prices — which is why strongly consistent protocols like Paxos resist scaling even inside a datacenter. So the platform scales by segmenting, not consensus: the bus fans work across fair lanes, the memory answers caller-side, and only the ledger pays for agreement. Tolerating attacks: a denial of service is not a partition and cannot be modeled as one, so it needs a different posture — the honest surface, where degraded states are rendered rather than masked and every answer stamps its provenance. Mobile wireless: unpredictable loss and proximity-first applications, answered by going close — the tunnel carries the surface to the reader, Tigris serves retention near the requester.

The three pressures share a root, and seeing it is the point of the dive: each is a place where the safety/liveness trade recurs under a new kind of unreliability — growth-coordination, malicious disruption, wireless loss — and each is met not by defeating the trade but by re-spending it. Scale is bought by refusing global consistency; attack-tolerance is bought by refusing to lie about health; mobility is bought by refusing distance. None of the three escapes CAP, and none pretends to. The paper names branded component systems as the future that copes with all three, and this platform is one such system — its single answer, threaded through parts that are segmented, honest, and near, is the same branded identity the rest of the course follows.

The three futures look unrelated — a capacity problem, a security problem, a physics problem — but the paper places them together for a reason, and naming it is the dive's whole move: each is a fresh site where the safety/liveness trade recurs under a new kind of unreliability. Scalability adds growth-coordination as the unreliable element — the more resources you coordinate, the more the coordination itself is exposed to the trade, which is why even a partition-free datacenter struggles to scale strong consistency. Attacks add malice — a denial of service is a form of unreliability the message-loss model never captured, so the honest-surface stance treats it as something to disclose rather than something to mask. Mobility adds wireless loss and motion — partitions are rarer but loss is constant and geography is destiny.

The platform's answer to all three is the same shape: re-spend the trade, do not try to beat it. Scale is spent by declining global consistency and keeping the consistent core as small as correctness allows. Attack-survival is spent by declining to fake health, so the surface's degradation is a truthful CAP stance the reader can act on. Mobility is spent by declining distance, pushing service and retention to where the reader stands. That the paper files this future under branded component systems is not a coincidence this course leans on lightly — it is the thesis: the identity that lets a system segment its trade is the same identity that lets it stay coherent while it scales, degrades, and moves. One thread, three pressures, one standing system.

The pressures rule — the three futures the paper names are three platform stances: segment to scale, disclose to survive attack, go close for mobile. Each is the safety/liveness trade recurring under a new unreliability, and each is met by re-spending the trade, never by escaping it — held together by the one branded identity the paper calls the future.