« February 09, 2004 | Main | February 11, 2004 »
February 10, 2004
ETCon 2004 and DDTI Photos
I've posted some photos from the Digital Democracy Teach-In and the Emerging Technology Conference to my photo gallery. This was the first time I used Zach Wily's iPhoto2Gallery plug-in for iPhoto and my first real use of iPhoto 4.0. The plug-in worked great and made uploading pictures a breeze. iPhoto 4.0 is much faster and didn't hang at all as I used it. All in all, a much improved experience.
05:57 PM | Recommend This | Print This
Kill Apps for Your Cell Phone
Rael Dornfest and others are talking mobile hacks. There was lots of fun things, but here's a few killer cell phone apps I didn't want to lose track of:
- Opera - a real browser for your phone.
- Agile Messenger - a better IM client for your phone.
- miniGPS - location-based alerts and messeging for your phone.
For more info, you might try MobileWhack or Howard Forums.
05:23 PM | Recommend This | Print This
ETCon 2004: Robert Kaye on Social Networking-Based File Sharing Networks
Robert Key advocates Bluetooth this year
|
Robert Kaye (slides) is describing social network file sharing systems. The primary purpose of the social group is to share, discover, and protect network. He proposes a hierarchy or tribes, chiefdoms, and states with leaders at each level and "tribal elders" who set the policies about who gets in. This sets the trust network.
Strict policies limit search horizons, large search horizons limit security. Social networks can grow quickly. The trick is to balance growth and security. Weak ties from social networks are stronger that random P2P connections. The networks lets you explore your strong and weak ties who are more likely to share your tastes and interests that strangers. Social network-based file sharing systems trade off quantity for quality.
Discovery includes ratings and recommendations from people in your social network to help guide and strengthen search results.
Robert likes a two-part system with a central server architecture. Algorithms for authentication are difficult to do in a decentralized system. The central server offers a Web service interface that allows user to build any kind of social network application. The server has no knowledge of what clients are doing and thus clients are protected. The central server also solves the P2P bootstrapping problem of how to get clients linked-in.
For the client, you can start with an open source client like Gnutella, although the standard routing query protocol is problematic, so something like the canonical identifiers from Bitzi, MusicBrainz, and IMDB's distributed hash table-based content searching system. Include the BitTorrent swarming P2P system for file transfer.
Using the network requires and invitation, port changing, and make sure everything goes over an ssh tunnel. Although, if you're going over ssh, you don't need an ssh tunnel since no one can tell what's going on anyway (except maybe through traffic analysis).
When the network is compromised, the attacker can see the same thing that a single network can see. So the damage, depends on how powerful individual nodes are. That goes back to the policies set by the "tribal elders." The idea is that this a toolset for building social networks, not the network itself and the elders use the toolset to create a network that follows their policies. When two tribes trust each other, they can connect and interact, expanding the search horizon.
Robert recommends "Guns, Germs, and Steel" by Jared Diamond, "Smart Mobs" by Howard Rheingold, and "Urban Tribes" by Ethan Walters.
02:38 PM | Recommend This | Print This
ETCon 2004: Fluid Time: Scheduling Washing Macines
I went to a talk on the Fluid Time Project by Molly Steenson and Michael Keislinger (slides [PDF]) . Some interesting social findings on how time works in groups, but the most interesting part to me was the discussion of instrumenting washing machines in a student housing project and then providing scheduling for them. The system negotiates schedule changes, sends alerts when the laundry's done that depend on how busy the machines are and how prompt the person usually is, and lets users check the status of the machine using their phone. A long time ago, out of frustration, I wanted to build a similar system for scheduling car repairs and doctor's appointments. This is that system.
12:40 PM | Recommend This | Print This
ETCon 2004: Dave Sifry on Technorati
Dave Sifry talks up Technorati
|
I'm at the session being done by Dave Sifry, creator of Technorati.com
- Over 1.6 million sources tracked
- 11,000 new weblogs created everyday, up from 4-5K per day in March 2003.
- About 35% of weblogs are abandoned (no posts in 3 months)
- Over 100,000 updates per day.
- Median time from weblog post to live index (on Technorati) is 7 minutes. This makes the engine usable for tracking weblog conversations.
The nice thing about Technorati is that it tracks deep links. Almost no one links to www.amazon.com. They link to some specific page on Amazon (which, BTW, Amazon has enabled by having a RESTful architecture).
Dave points to a things he hacked together last night to point to products. Dave asks for an experiment. He asks the audience to link to the product page and then periodically check the cosmos for the page to see when they links appear on Technorati.
Dave talks about the Power Law of Blogging and shows his data. The data shows that when you have fair access to media, there will be a relatively small number of things that are linked to by a lot of people. When there were only three networks, they were all even distributed. When there are hundreds, you get a power law graph.
Technorati as Platform Dave's commitment:
- XML API for all functionality based on a RESTful architecture that is free for non-commercial use.
- Today: Link Cosmos, keyword search, top 100, breaking news, and current events.
- Perl, Python, Radio, C#, and ASP interfaces (see developers.technorati.com)
There are IM/SMS notification, movable type plugins, threading on weblog readers, and a high priority indexer (using the high priority indexer). Some application directions: open reviews (RVW format), keyword and Cosmos filters, discovery and filtering of subscriptions lists, vote links (differentiating between links as endorsements and links to non-endorsements), and geographic search and indexing.
11:16 AM | Recommend This | Print This
ETCon 2004: Tim O'Reilly Keynote
What's on Tim's Radar:
The net is the platform. The new killer apps of the Internet, eBay, Yahoo!, Amazon.com, PayPal, Mapquest, and others are on the O'Reilly best seller list, but moreover, are running on a new platform called the Internet. The software lives somewhere other than your local machine. These apps run on open source, but themselves are not open source.
Tim compares Amazon and Barnes & Noble. Amazon has harnessed the user community, but B&N has not. Amazon outsells B&N. Mapquest has not offered any social services. GeoURL and GeoAnnotation are some interesting projects in this area. Microsoft's Mappoint has an API, but they have no clue about end user participation. Tim believes that the first mapping site to bring social aspects to their site will come out on top.
Wordspy is a web site that creates dictionary style entries of words as they are coined. The site provides insight to culture as it evolves and words in action.
iPOD is a combination of hardware and backend system. However, wireless is missing. Also, there's no architecture of participation. GarageBand let's you create music, but there's not way in iTunes to share it. iChat knows about buddies, but iTunes and iPhoto doesn't. Orkut let's you organize friends into groups. More sites need to use this kind of information.
Managing relationships. Identity is about more than who I am and who you are, its about how we relate. Where is my address book? Is it on my phone, on my PC, on PayPal, on Okrut?
Why can't we have P2P identity sharing? Why can't I manage my friends on a local level and tie that into this other applications? (This is the DigitalMe idea in some ways.)
On the issue of mining the net for data, Tim talks about Technorati. He also mentions OrgNet.com's analysis of book purchase patterns. The study revealed a divided populace. Only two books connect the liberal and conservative book buyers. Interesting.
Here's Tim's summary:
- Hacking in all its forms
- Second generation network effects
- social software
- network enabled market research and data visibility
- Architectures of participation
- Getting beyond single device
- Robotics an hardware hacking