BarCampJozi 08 day 2

An even more interesting day at BarCampJozi. It’s been really great meeting everyone! My twitter account is much fuller now.

As before, these are just a few notes and impressions of mine, mostly things that interested me.

Human quantum interaction

Talked about some experiments into ESP and psychokinetics. Mentioned first the human expectation effect (eg. placebo effect).

Some comments from the audience on mirror neurons and very subtle body language cues, and some scepticism on systematic errors and the other usual objections. Tough crowd.

Some discussion on amplification of quantum events through neural networks. Basically, if we are willing to throw away causality, then anything can happen. No surprise there!

Systems thinking
Carl Spies

Speaker interested in how people learn and make decisions, work out what things are right, etc.

Captology (Computers as Persuasive Technologies): how to use computers to make people make better choices, how websites can be made to sound sincere. Graphed the data-information-knowledge-wisdom progression as a positive slope on a graph with axes “understanding” and “connectedness”. Systems thinking is a meta-discipline for describing understanding, pushing us along the path to wisdom.

Ages of humanity: the most important things have moved from “matter” (agrarian) to “energy” (industrial age) to “information”.

Biomatrix: a cross-weaved net. Strands are “actions” or “processes”, and intersections are “identities”. Changing identity needs to look at all the actions involved.

Bluetooth security

Problems with bluetooth security: computational complexity of cracking has greatly decreased, and some devices have hard-coded default security pin codes.

Some discussion of handover from tower to tower, and how it maintains IP address — except when you handover to a different cellphone network. Also picocells, which one can buy and associate with a cellphone network to extend a network. Also can be used to fool cellphones into associating with picocell, and thereby leaking all sorts of data.

Much discussion on mesh networking and community networks. Legal challenge if they undermine cellphone network revenues or interconnect multiple entities. Interesting project in Orange Farm:; blogged about more fully here:, and a FAQ:

The legal framework for phone tapping in South Africa: govt and service provider each have a piece of hardware. Both need to be enabled, with court order (related to national security) to get service provider to do so. Tapped signal only available on govt side. Sounds good!

Social media

Stressed that people need to use their online social networks in the same way that real-world social interaction happens. Applies to corporates using social networks too — eg. in marketing, if the main purpose of the interaction is to sell, you lose. You need to focus on the interaction itself, and giving value to the person too.

Some partial rebuttals: people use networks in different ways, and in particular have different contexts in mind — this affects whether they will view favourable requests from people they don’t know, or corporates. Another point: the communication medium matters too: an SMS with just the word “Thanks” doesn’t have content, and is an interrupt.

Another good point: values and norms differ around the world, but almost all online technology is built around the values of Silicon Valley — not always right!

XMPP and social network architecture
Blaine Cook (formerly chief architect of Twitter)

Starts with a brief history of physics, leading to the start of the web at CERN, as a way of organising data, and work at UC Irvine designing the basis of the internet and web paradigm. BUT problems, eg., Twitter fail whale. Problem: polling the server is inefficient, as most responses are just, “no”. Root problem is that http is a pull protocol, not good for social networks.

XMPP (formerly Jabber) interesting system:
– persistent (ie. saves on initiation)
– lightweight
– bi-directional
– asynchronous
– guaranteed server identity
Now what?

Extending REST for decentralised systems — do subscription / publish system: Jabber PubSub. Nice examples using some politically-loaded comments about current US election candidates. Now outlining a pseudo-Twitter system, where posting messages gets sent as XMPP requests to everyone who’s interested.

Talking about work on FireEagle, a location-aware system. Vital aspect: latency. If it’s low, can send messages like, “you’re walking past your friend right now.” But high latency, ie., http, is too slow for this. Messaging is good, also, for tying together systems, eg., Facebook and Flickr for new notifications. Also, Jabber identity now starts to serve as a single ID.

100% reliability and ordering are LESS important for social networking applications, so not a key constraint here. Some discussion on caching and trunking — the idea at this stage is to use the feeds between servers and services, rather than last mile.

Very interesting discussion about different sorts of applications, and perhaps embedding XMPP in the browser, etc. I’m not writing it all up though.

Engineyard — making cloud management systems. Now rewriting in ActiveMQ or rather RapidMQ (protocol AMQP). Probably a better system to tying together servers (ie. middleware), rather than human-centric notification type events.

Project Diaspora

Talked firstly about his personal experience, as a diaspora member, of helping his siblings through school. The breakthrough idea was that remittances can just vanish into a hole, or can build lasting change. Aim of project is to harness the diaspora for powerful, focused projects. Money needs to benefit communities rather than just individuals. It’s currently hard to find lasting projects that have come from aid — mostly aid builds dependency. Diaspora more likely build self-sustaining, well-thought-out, efficient projects.

Diaspora remittances are now around half of aid sums going into Africa, and growing. Rough figure 2007: $39 billion.

Aim: mashup of social network site, remittance processing, Ushahidi for mapping projects, etc.

Some physics
Paul (yep, that’s me)

Discussion about the Large Hadron Collider, the Higgs boson, eternal inflation and the possibility we’re a small part of a larger universe at higher vacuum energy. The most interesting idea for some people is that there’s no reason to prefer the universe either having or not having a beginning.

Behaviour Driven Development

Awesome talk (great pictures) on the progression of QA testing to regression testing, unit testing, test driven development, and finally behaviour driven development. In practice, similar to unit testing, but written in a way more descriptive of behaviours. Thus tests describe the system’s functionality, and so can be even given to end client to show what the system does and does not do. This topic requires a much longer description than it will get from me here!

Last night was also great, more details in the most post. Tomorrow I start MobileActive 08, so three more days of conference! Hopefully I’ll be able to get some live-ish blogging going — I expect there will be many very interesting, and not as geek-focused, talks.