From dc3460db260ac01c96600dd685ce4a08fa1f87e8 Mon Sep 17 00:00:00 2001 From: Jeff Burdges Date: Thu, 30 Aug 2018 14:31:11 +0200 Subject: [PATCH] Start a timing section --- NetRNGs.md | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) diff --git a/NetRNGs.md b/NetRNGs.md index 791db31..3525934 100644 --- a/NetRNGs.md +++ b/NetRNGs.md @@ -113,14 +113,14 @@ numbers we might argue can be used like collaboratively produced ones. Alistair says we cannot wait for finalisation here, but disagreement results from multiple chains running in parallel. -### Algorand ?? +### Algorand Algorand has similarities to Tindermint, but selects validators randomly from a pool, instead of using a fixed set, and does a Byzantine agreement for each block. All nodes produce possible blocks and a VRFs helps choose among them. -### RANDAO ?? +### RANDAO Nodes commit to a hash onion and use the layers like a local VRF, revealing one layer whenever doing so allows them to produce a block. @@ -181,3 +181,22 @@ Are there other fishermen related uses? Rob mentioned ordering fishermen, which sounds problematic. +# Timing + +TODO: Comment on relative block time vs. more absolute synced clock times. + +### RANDAO + +We noticed possible timing issues with RANDAO that impact most other +schemes: An attacker can manipulate the timing of their blocks to +help maintain a forked chain. TODO: Add details. + +### Ouroboros rate metering + +Ouroboros paper 3 describes minimum rate requirement for protection +against long term attacks. The idea goes that a long-term attacker +starts with relatively little stake, so initially they produce blocks +slower than the overall network, but as their stake increases this +rate goes up. Ouroboros wants to prevent this initial slow start, +but this requires +