Con-D.2.7 — Garbage-Collection Analogy Backfires: Modern GC Moved Away from Stop-the-World#

Severity: D (Substantial) | Sphere: Se1, Se6 | Target: ax25

Reply 1b’s efficiency argument for periodic over continuous redistribution uses the analogy of stop-the-world garbage collection (periodic Jubilee) vs. real-time garbage collection (continuous taxation), claiming periodic GC has lower amortized cost. Taken seriously, this analogy argues against a Jubilee system:

  1. Stop-the-world GC is universally recognized as inferior for most applications. The JVM abandoned pure stop-the-world GC in favor of concurrent collectors (CMS, G1, ZGC, Shenandoah) precisely because stopping the world creates unacceptable latency spikes. Modern GC research focuses entirely on minimizing stop-the-world pauses.

  2. The correct analogy for modern GC is progressive taxation. Modern generational GC distinguishes between short-lived objects (collected frequently, cheaply) and long-lived objects (collected rarely, expensively) — precisely the structure of progressive taxation.

  3. Anticipation effects destroy the Jubilee mechanism. Economic agents have agency (ax15!) unlike software objects. Rational agents anticipating a Jubilee will convert resetable assets to non-resetable forms, move wealth offshore, consume capital, under-invest before and over-invest after the reset. This is the Lucas critique (1976): agents’ behavior changes in response to announced policy, potentially rendering it ineffective or counterproductive.

  4. The “constant balancing” overhead argument is empirically wrong. Countries with strong continuous redistribution (Denmark, Sweden, Norway) have simpler regulatory frameworks than countries with weak redistribution (US). The US’s complex regulatory apparatus is a product of inadequate redistribution, not excessive redistribution (Kleven 2014).

Academic support: Lucas (1976), Carnegie-Rochester Conference 1:19–46; Tene (2005) on GC evolution; Kleven (2014), JEP 28(4):77–98.

(Source: C2.7 from OOv1 Critique Round 2.)