Blog
How we build backend technology for MMORPGs.
Netcode in Hanamori, part 1: the tick, the wire and the area of interest
The 20 Hz tick, SBE frames with full and delta updates, who is told about whom, and the server tick every entity update now carries.
Netcode in Hanamori, part 2: predicting yourself
Prediction on a shared float32 step, the input ring buffer, the server input queue, and reconciliation by replay: snap above 4 m, ease below.
Netcode in Hanamori, part 3: seeing everyone else
Tick-stamped interpolation, the clock estimator, and an adaptive delay with worked jitter and stall examples. Why a fixed 100 ms left no margin.
Netcode in Hanamori, part 4: hit grace and the edge cases
Round-trip measurement, monster hits that wait out your ping, the trace ring, what is not rewound, and a catalogue of what goes wrong.
Hanamori on Rune, part 1: topics, views and boot
Part 1 of 3. What a Rune topic, a view and a transaction are, how the 14 topics are laid out, how a server proves it has caught up at boot, and how a player logs in. Real types, code and diagrams.
Hanamori on Rune, part 2: positions, rewards, loot and portals
Part 2 of 3. Positions that cost freshness instead of memory, two-record inventory moves, the 8-op loot pickup and portals that expire, each with the code and a diagram.
Hanamori on Rune, part 3: the realm handoff
Part 3 of 3. One hero from the meadow into a dungeon and back across two servers that never connect: ownership records, tickets, the return barrier, fencing and every failure branch.
Shared-nothing: how Hanamori's servers scale without locks
Every thread owns its data, and ring buffers are the only way work crosses between them. How Hanamori's meadow and realm servers are built that way.