Registries After Dark, Part 2: Distributed Random Access Merkledags
summary
Builds a RISC‑V emulator in Rust, compiles a tiny C program with a RISC‑V cross compiler and translates each store into an OCI image layer so each memory write becomes an 8‑byte OCI layer pushed to Docker Hub under tags that are memory addresses. Retrieval uses crane to fetch tags and hex dumps show the exact bytes. It proves registries can hold program memory and the demo works but is painfully slow, wasteful, and a poor fit for registry DAG semantics.
AI-written summary · about
chapters (7)
- 0:00Start
- 2:02A General Model for Computers - Computers are constructed from components. . Physical components are very different from one machine to another. . Despite the heterogeneity, most adhere to a logically sinilar rodei.
- 3:51von Neumann and Stored Programs - Comon pattern developed of using paper tape (and later punch cards) to program machines
- 5:19von Neumann and Stored Programs - Common pattern of using paper tape (and later punch cards) developed to program machines - Stored Progran Concept: instructions and data can both be stored in menory.
- 5:59Instructions as Data - The definition and encoding of instructions is known as an Instruction Set Architecture (ISA)
- 31:32Why? - Everything is bits, and bits can be stored and retrieved - Distribution: how easy is it for folks halfway around the world to access memory from the progress I's debugging? • "Free": registries are widely available and relatively cheap.
- 32:40The network is slow. . Do you need distribution? • Does anyone really need to access the memory of a process I'n debugging? - Are you taking advantage of the data structure? • Is this content useful for structuring
related talks
05mRISC Zero: Boundless Execution on Solana
51mRibbit with Memory Morphisms
49mFull Rundown
41mCoding like it’s 1977: Ruby on the Apple ][
45mInteraction with real-world JavaScript from compiled Rust
21mOpen Standard Software Acceleration for RISC-V
33mThe Runtime Rodeo; Where Open Source Image Behavior Is Tamed
17mRunning WebLLM in the Browser on RISC-V Toward Lightweight, Local AI Experiences
