Think I'm having my "marble diagram" era. Latest experiment: are database isolation levels more easily explained using marble diagrams?

Still doing Gossip Glomers. The broadcast code laughs in the face of Maelstrom's `--nemesis partition`.

But the latencies... Median and 99th percentile are way below the exercise thresholds. Last percentile is horrendous. Feels like hunting down that edge case will be "the other 90% of the work".

Marble diagrams are my favourite part of the RxJS ecosystem. But for distributed systems I've mostly seen sequence diagrams. I wonder if anyone's tried marble diagrams instead?

The Gossip Glomers challenges escalated rapidly with #3 "broadcast". It introduces inter-node communication and I realised that A Tour of Go didn't include anything about debugging.

Gossip Glomers #2 "unique id generation" was nicely setup. I hadn't read ahead (inter-node communication appears later) and the maelstrom config for the exercise assumes identical nodes. Realising the setup means generating unique ids without coordination leads to the simple answer (uuid).

Having skipped Advent of Code, in Jan (new year's eve totally counts as Jan) I'm using the distributed systems challenges from fly.io / Jepsen to learn Go. So far it feels more like learning Maelstrom than learning Go.

Were Xerox Corporate Internet (CIN) and Xerox Network System (XNS) the same thing? I'm reading some Xerox papers from the 1980s. They bounce between the two names, but never within a paper. XNS seems later?

Huh. Paul Brian Koski's 1987 thesis "Examination of a Directory Service" is behind a paywall in 2024.

Subject of Boxing Day's reading is Xerox's Clearinghouse system for finding "named objects in a distributed environment" (1980s, I think?)