Skip to content

Instantly share code, notes, and snippets.

View ept's full-sized avatar

Martin Kleppmann ept

View GitHub Profile
ept / micromerge.js
Last active November 15, 2021 12:59
const ROOT_ID = '00000000-0000-0000-0000-000000000000'
* Miniature implementation of a subset of Automerge. Supports the following:
* - only map, list, and primitive datatypes (no table, text, counter, or date objects)
* - no undo/redo
* - no conflicts on concurrent updates to the same field (uses last-writer-wins instead)
* - no API for creating new changes (you need to create change objects yourself)
* - no buffering of changes that are missing their causal dependencies
* - no saving or loading in serialised form
datatype 'val msg
= Propose 'val
| Accept 'val
ept /
Last active December 7, 2018 10:15
Notes on Mark Hayter's stylus talk

Stylus input and the benefits of a standard

Notes from a talk by Mark Hayter

Mark Hayter is Senior Engineering Director in the Chrome OS Hardware team at Google, and currently a visitor from industry with the Computer Architecture team in the Computer Lab.

  • 1990s: Apple Newton, Palm Pilot used graffiti input (one letter at a time, stylised), Lectrice
  • 2000s: Microsoft Tablet PC

Notes from Coed:Ethics conference, London, 13 July 2018

Cori Crider - "When data kills" / Weaponised AI

  • Imam Salem bin Ali Jaber preached in Yemen against Islamic extremism; guest at a wedding of relative; hit by a US Hellfire missile fired from a drone. Relative Faisal made contact with Cori, went to Washington DC. No explanation ever made by government (although compensation was paid).
  • Decision-making process behind the attack not known exactly. But there is significant evidence that such attacks serve to further radicalise people; attack results in ~3x more new recruits than extremists killed by attack.
  • Most drone attacks are not on named individuals, but rather "signature strikes" — a euphemism for killing people the military doesn't even know, but who match a certain behavioural pattern (perhaps based on metadata — Hayden: "we kill people based on metadata"
  • Skynet (known through Snowden relevations): use machine learning to try to find courier
import org.apache.avro.Schema;
import org.apache.avro.generic.GenericData;
import org.apache.avro.generic.GenericDatumReader;
import org.apache.avro.generic.GenericDatumWriter;
import org.apache.avro.generic.GenericRecordBuilder;
ept / dataloss.rb
Created January 24, 2017 22:17
Calculate the probability of losing all replicas of a partition in a cluster
# Parameters:
prob_nodefail = 0.001 # Probability of a single node failing
replication_factor = 3 # Number of copies of each partition (r)
partitions_per_node = 256 # Number of partitions per node
max_nodes = 10000 # Maximum number of nodes to consider
# (n - r)! / n! == r! / (n choose r)
# Intuitively: the fraction of the n! possible permutations of n nodes that
# results in the replicas of one particular partition to be mapped to three
# particular nodes, in a particular order.

TRVE Data update — November 2016

This update was sent by email to the TRVE Data mailing list on 1 December 2016. Please subscribe to the mailing list or follow us on Twitter for future updates.

Hello TRVE Data supporters,

Here's an update on what we've been doing in November.

  • Stephan has made great progress on understanding Tor network traffic. He is now able to trace almost all data transfer through the various layers — application data, Tor circuits, calls to the OpenSSL library, down to actual network packets — and thus analyse each packet to determine why it was sent or received. This is especially important when using Tor on mobile devices, since every network packet uses battery energy and costs money (depending on your data plan). By labelling network packets, we can compare protocol changes to
ept /
Last active October 31, 2016 20:31

TRVE Data update — Sept/Oct 2016

Hello TRVE Data supporters,

Sorry for the lack of an update last month. Here's what we've been up to in September and October.

Diana and Martin presented an introduction to end-to-end encryption protocols at the Strange Loop conference in St Louis, and repeated the talk in Cambridge. We illustrated network communication and man-in-the-middle attacks by passing pieces of cardboard back and forth on the stage -- a visualisation that went down well with the audience.

Alastair and Martin have been working together with Dominic Mulligan and Victor Gomes, also researchers in Cambridge, on formalising our JSON CRDT definition using the Isabelle proof assistant. We are converting the hand-written convergence proof in our paper into a machine-checkable proof,

TRVE DATA monthly update: May 2016

Hi everyone, the month of May has flown by, and it's time for our second monthly update for the TRVE DATA project.

In case you missed it, last month's update is here:

Here's what's new in May:

  • Stephan has been doing some measurements to determine the overhead of running a Tor hidden service on a smartphone, which could be used to exchange messages between devices directly in a privacy-preserving way. He found that just keeping the hidden service alive produces several hundreds of megabytes per month in traffic. To a large extent this seems to be caused by regularly downloading information about the status of all Tor relays. As this amount of overhead is not acceptable for many users with a limited data allowance, an open question is whether we can reduce the amount of data that needs to be downloaded, and what the trade-offs are.
ept /
Created April 29, 2016 15:26
TRVE DATA monthly update: April 2016

This is a copy of an email sent to the TRVE DATA mailing list.

TRVE DATA monthly update: April 2016

Hi everyone, thanks for subscribing to the TRVE DATA mailing list! We're going to try to build a habit of posting a monthly project update here, and this is the first one.

For background, we published a blog post explaining the rationale and goals of the TRVE project. And here's a little video showing our collaborative text editor prototype.

Today I'd like to briefly introduce the people in Cambridge who are working on various aspects of TRVE: