This is the third in a series of posts which represent my notes, impressions, and in some cases audio, for the sessions I attended at the 2007 Web 2.0 Expo.

Writing Voice Mashups with Mechanical Turks and Maps (Thomas Howe, Thomas Howe Consulting). A great sample mashup application, which integrated telephony and web services to provide providing a better experience for patients who need healthcare assistance after hours. To me, the fact that the application was a mashup was really of secondary importance. More important was that Howe showcased really eye-opening capabilities that we now have available as application developers:

1. Amazon’s Mechanical Turk. In this application, Howe used Turks in two cases: qualified nurse-Turks to screen the incoming calls and determine which needed to speak with a doctor right now; and transcribers, to turn the patient’s incoming call into text that can be attached to their permanent medical record. Howe echoed Bezos’s theme of “artificial artificial intelligence,” emphasizing that we can now tackle practically any problem, even if it takes human reasoning as part of the process. Interesting annecdote: Howe’s mashup Turk-tasks were paying $3.00 a pop, which it turns out is a really high Turk fee; most current Turk tasks pay pennies per. He said the Turk community was all over his task postings like “stink on …” based on this “high” fee.
2. Telephony+web in a seamless, simple-to-build application. Interactive voice response (IVR) applications have been around for years, but they’ve been standalone apps, running on expensive standalone hardware, and requiring special off-the-beaten-path expertise to program. Howe shows us how the combination of telephony service providers and a new standard called VoiceXML enable us regular web developers to create custom IVR+web applications really quickly and without the need to invest in special hardware.
3. SMS as a workflow confirmation tool. When the incoming call was from a cellphone, Howe’s mashup issued a series of SMS messages that keep the patient updated as to the status of their request: “A nurse is now reviewing your message,” etc. Howe’s app used a StrikeIron webservice to send its SMS messages; there is a per-message fee but StrikeIron claims to be able to reach virtually any SMS user on any carrier worldwide through their webservice.

All in, really cool stuff, opens up a lot of possibilities without requiring dedicated equipment and specialized skills. mp3 audio (starts a few minutes into the presentation)

Tuesday Keynotes

• Launch Pad #1: Dmitry Dimov, Cofounder & Product Chief, and Brian Mulloy, Co-founder and CEO, Swivel. A pretty interesting web-based service that allows people to “upload and explore data.” “Half the internet is missing,” Mulloy says; the web is good and text, images, videos, etc., but misses the boat on data. Swivel data is public; enables open conversation around key issues backed up by data, statistics, analysis.
• Launch Pad #2: Luke Sontag, President, Vidoop. A novel visual authentication approach that does away with passwords in favor of a visual, photo selection interface. A bank of ~9 images is presented, each with an assigned letter code, each falling into a category such as “airplane,” “boat,” “computer,” etc. The user knows their pass-sequence is “boat, food, tree” and they locate the photos in these categories and enter the letters associated with these photos. So this is un-phishable; there is no fixed password to be phished. Apparently compatible with OpenID. Cool idea, though I am a little skeptical of his business model, which is that advertisers would place images of their products to be show in the image bank used for login.
• Launch Pad #3: Mike McCue, President, CEO & Co-Founder, Tellme (now Microsoft). Says their goal was to bring the benefits of the internet to the telephone. For developers, Tellme designed an open application platform, which enables us to build apps that target any phone; in other words, interactive voice response apps, see my comments about Thomas Howe’s presentation. Tellme evidently developed the VoiceXML standard, and they have a development environment available at studio.tellme.com. For end users: the built a really slick “voice portal” that enables any phone to call up and get news, sports, stock quotes, restaurant reviews, and perform business searches. He dialed the service during the demo (the number is 1-800-555-tell), asked for “business search,” was prompted for location, said “San Francisco, California,” was asked what kind of business he wanted, said “ice cream,” was prompted for neighborghood, said “Union Square,” and the IVR app began reading off ice cream parlors nearby to Moscone West; he selected Coldstone Creamery on Ellis Street, and it ready off a phone number. Really nice, especially for free. If you call from a mobile device, it will send text messages with the results; and they also now have a local phone client app (not sure what platform). Cool.
• Mobile 2.0 (Ajit Jaokar, CEO, futuretext; Mike McCue, Tellme; Ilkka Raiskinen, Nokia; Paola Tonelli, Vodafone Spain). Pretty much a fluff piece; O’Reilly needs to be careful about involving most old-line companies in this kind of conference.
• High Order Bit: Architecture for Humanity (James Baty, Sun Microsystems). A good cause and all, but only marginally related to the conference. Shouldn’t have been on the agenda.
• State of the Web 2.0: Measuring the Participatory Web (Bill Tancer, Hitwise). Some interesting stats. 3400:1 visit ratio, Wikipedia versus Encarta websites. Measured percentage of uploaders versus viewers: surprisingly low (.16 % for YouTube .2% for flickr), with wikipedia far exceeding these, at 5%. I loved one of his demographic categories: “Shotguns and Pickup Trucks” … These were fresh stats, and I think they need to do more analysis, as always some initially unexpected user behaviors might be easily explained: for example, might the fact that it takes a really long time to upload a video on YouTube explain how comparatively infrequent this act is?
• High Order Bit (David L. Sifry, Technorati). Growth of “active” blogs starting to level off.
• Eric Schmidt in Conversation with John Battelle (Eric Schmidt, Google; John Battelle, Federated Media). Schmidt announced that Google will soon add a presentation tool to the Google Docs suite. Battelle asked Schmidt why Google would want to buy DoubleClick with their “punch the monkey” style banner ads, got a chuckle. An even bigger laugh when Battelle asked about Microsoft and AT&T stirring the pot about this acquisition, saying “antitrust! antitrust!” Schmidt made a funny face and said, “Who was that? Did you say Microsoft? AT&T? Antitrust?” video of the Schmidt interview on viddler
• mp3 audio

Case Study: Digging into the Technology Behind the Development of Digg (Owen Byrne, Senior Software Engineer, Digg). Fun to get a behind the scenes look at how Digg got started and the growing pains they went through. Starve as long as you can, even if you do plan to get financing eventually; they survived on one server for a really long time. Today: 100+ servers total, perhaps half that are true production machines. One of the steps they had to take to support traffic growth: denormalize all database tables. Just recently replaced MySql text indexing with Lucene: a win for Java mp3 audio slides

Rich Internet Application Platforms (Ryan Stewart, Threecast; Ben Galbraith, Ajaxian; Jeff Mancuso, Magnetk; Chris P. Saari, Yahoo; Dirk-Willem van Gulik, Joost). Of the bunch, van Gulik was the most interesting to listen to. We had heard from our Cocoon community friends that the Joost guys were uncompromising when it came to user experience, and Dirk confirmed that attitude, explaining why they chose the Mozilla platform and why something like Adobe Apollo wouldn’t have provided them enough control of the user experience. mp3 audio

API and Mashup Best Practices (John Musser, ProgrammableWeb.com). Interesting survey of different API approaches. mp3 audio slides

Next up: My Web 2.0 Expo: Wednesday (including lots of audio)

This is the second in a series of posts which represent my notes, impressions, and in some cases audio, for the sessions I attended at the 2007 Web 2.0 Expo.

Monday was my first full day at the Expo, and the day I had to face the sad truth that I could only attend 12.5% of the sessions available, since there were eight simultaneous tracks. It is a current meme that having too many options leads to unhappiness; and I have to say it was true for me, for a while at least. But in the end, I just had to pick one session and try hard not to regret the seven I was missing.

All You Need to Know About Django (Adrian Holovaty, Django). Adrian is a good guy and it was great to hear the story to Django from the horse’s mouth. I came away, though, unconvinced that we should abandon our current favorite platform Ruby on Rails in favor of Python / Django.

Rich Internet Applications with Apollo (Mike Chambers, Adobe). Mike got burned by the failure of the wireless network, which preventing him from demoing, which pretty much killed this presentation. Ironically, Adobe was the sponsor of the wireless network [1] I came away from this session totally underwhelmed with Apollo, wondering why anyone would want to use it. A later demo by Kevin Lynch during the afternoon keynote changed my mind.

From Software to Webware: How Web-based Applications Will Shake the Software Industry (Panel: Paul McNamara, CEO, Coghead, facilitator; Ismael Ghalimi, CEO, Intalio; John Seely Brown, Senior Fellow, Annenberg Center at USC; Vishal Sikka, CTO, SAP; Martin Wegenstein, Former CIO, Autodesk). Note to self #1: avoid panels. Note to self #2: avoid presentations by old-line companies trying to look cool by hanging out at the Web 2.0 conference. There was one bright spot: John Seely Brown, but man, he looked uncomfortable up there with a panel that included the CTO of SAP, a (former) CIO, and a company that sells SOA tools to the Global 1000. My favorite JSB quote (paraphrased): “I may be the only one on the panel today that holds the unpopular opinion that IT departments generally do more to impede progress than to promote it.” Zing. JSB also talked about collaborating users having a “gaming disposition,” that is, the kind of we’ll-pull-it-together-ourselves mindset that a World of Warcraft guild leader has, “assembling resources towards a goal.” My desire not to miss a good JSB comment was the only thing that kept me from running out of the room screaming.

Tagging that Works (Thomas Vander Wal, InfoCloud Solutions). Great session, good insights around all aspects of tagging and folksonomies. I liked his term “pivoting” which he uses for the scenario where, as I drill down on a particular tag, let’s say “web2.0″, I pivot on another dimension, perhaps a related tag (”enterprise2.0″) or a different element of metadata such as a frequent user of “web2.0″ tags. Very funny demo sequence on Amazon.com, pivoting around the tag “talentless” … mp3 audio

Keynotes

• Conference Welcome (Tim O’Reilly). Nothing memorable.
• A Conversation with Jeff Bezos (Tim O’Reilly interviewing). Bezos’s opening comments focused on Amazon’s web services offering (he termed them “infrrastructure web services): Simple Queue Service (SQS), Mechanical Turk (”artificial artificial intelligence”), Simple Storage Service (S3), and Elastic Compute Cloud (EC2). He emphasized variable pricing, so no need to incur large up-front costs. More than 5,000,000,000 objects now stored in S3. 920,000,000 S3 requests on a peak day. 16,000 S3 requests on a recent peak second. S3 architected for “web scale computing.” Walked through an example application architecture for a video encoding app based on SQS, S3, and EC2. Showed how EC2 servers can be added on demand based on queue length. Talked about a spaceflight website that had extremely high peak demand when they posted video that got picked up by major news services, showed how use of S3 kept their bill low, just $200 for S3 services for the month that included this peak event. Says that EC2 is “completely capacity constrained” right now, he says “we’d much rather be demand-constrained,” trying to get there, but for now EC2 is invite-only. Web services “certainly not profitable today” for Amazon. O’Reilly pressed Bezos on why Amazon was straying from retailing in its web service offerings, who explained by saying “we looked around and asked ourselves, “what are we really good at?”
• Built to Last or Built to Sell: Is There a Difference? (John Battelle, CMP, moderating; Jay Adelson, CEO, Digg/Revision3; Joe Kraus, Co-founder & CEO, JotSpot, now part of Google; Mena Trott, President, Six Apart). Interesting point made that VCs don’t want returns of 5X or 10X — instead, they are looking for the home runs of 100X. So there is surprising pressure from initial investors not to sell for a 5X or 10X return.
• High Order Bit: Introducing Apollo (Kevin Lynch, Chief Software Architect, Adobe). Kevin did a great job explaining and demoing Apollo. Showed an eBay Apollo app currently in beta, provides very slick UI for on-line use, but also operates off-line; he used the example of putting his Web 2.0 conference badge up for auction, did it while disconnected from the net; the transaction queued up, and when he came back on-line, the queue was processed and the auction transaction uploaded. Kevin emphasized that Apollo apps are built using standard web technologies, such as HTML, JavaScript, AJAX, Flash/Flex. Their target for the Apollo runtime is 5-6MB, and is easily web-installable, which definitely sets it apart from other offline-capable solutions such as client-side Java and Lotus Notes. On the other hand, Apollo is a true client application, see by the OS as a native app. Targeting Windows, OSX, and Linux platforms, apps will run identically across all three. Apollo provides a sandbox that the application runs inside, does not provide full access to host OS, instead provides an API for things such as disk access. Showed a salesforce automation example built on top of salesforce.com APIs, again with offline capabilities; in this case, the ability to synch manage contacts and documents offline. Even more Lotus Notes-like, but based on familiar development tools, small, web-friendly.
• Launch Pad #1: Jay Bhatti, Co-founder, Spock.com. Amazing people search tool. Shows what tagging can really do, great tagging UI, nice “pivoting” capabilities (see my comments on the Tagging session). Still closed beta, too bad.
• Launch Pad #2: David Knight, Vice President, WebEx Connect. O’Reilly, what were you thinking? This was worthless.
• Launch Pad #2: Kerry Fleming, inpowr. A social network where people work together to set goals and achieve positive things in their lives.
• mp3 audio

Next up: My Web 2.0 Expo: Tuesday (including lots of audio)

[1] One of the conference people had the temerity to tell the Tuesday keynote audience, all 2000+ of us, that “the wireless vendor says everything is fine, we have plenty of capacity, so if you’re having a problem [it’s your fault], you should stop by the help desk [you idiots].” Square brackets are, of course, my interpretation of what she was saying. A little later she asked how many people were having wireless problems. About 1000 hands went up. Clearly the problem was in fact with the wireless setup; I suspect we were hitting limits on both connection capacity (max number of nodes that could connect at once) and bandwidth (I often had a sub-5KBps connection, sometimes sub-1000Bps). Well, it is no doubt a tough problem to solve, as this kind of problem seems to be more the rule than the exception at large conferences.

This is the first in a series of posts which represent my notes, impressions, and in some cases audio, for the sessions I attended at the 2007 Web 2.0 Expo.

I didn’t leave Chicago until 2:50 PM on Sunday, and then United decided to send my suitcase on a later flight … so it was actually a minor miracle that I got to any sessions on Sunday. What I managed to make was the second hour of “Ignite” sessions, which are a series of 5-minute talks presented by about anybody. Slides advance automatically every 15 seconds, so the pace stays brisk. Here’s what I saw:

• Justin Kan (Justin.tv) - The Justin.tv Launch: How to get a lot of press completely by accident and through no fault of your own. Interesting to get a first-hand perspective of Justin’s 24×7 videocasting-his-life world.
• Jane McGonigal (Institute for the Future, Avant Game) - Happiness Hacking An interesting concept, hacking one’s happiness. She references several books that might be worth reading.
• Andre Charland (Nitobi, RobotReplay) - Remote usability for the rest of us. Looks like a very useful and lightweight way to get a looking-over-the-user’s-shoulder perspective on how people use your website. Definitely on my must-try list.
• Andres Morey (Octopart) - South Pole Hacks. An interesting story: he worked on his Web 2.0 application while at the South Pole. But the two didn’t seem related really …
• Simon Wardley (Zimki) - Commoditisation and future stuff. He cranked through 70 slides in his 5 minutes, which was impressive. But for the life of me I can’t remember what he talked about …
• Nik Cubrilovic (Omnidrive, Techcrunch) - An Introduction to WebFS. Moderately interesting, although I am a bit skeptical of Nik’s primary points: that a standard will actually be supported among competitors in the web-storage space, and that a single standard will emerge that unifies both client application and webware storage.
• Kellan Elliott-McCrea (Flickr) - Casual Privacy. Great session, excellent perspective on the burden of privacy and a very useful suggestion for a lighter-weight, less-burdensome alternative. We’ve actually implemented something quite similar for a client, but Kellan takes the idea a bit further.
• Colin Bulthaup (Squid Labs, Potenco) - How do you create a power infrastructure in developing countries using human power. Great idea, interesting gizmo.

Next up: My Web 2.0 Expo: Monday (including some audio)

In the continuing thread of testing how far we can go towards reproducing rich client applications on the web, I’m currently working on a photo management application that is in need of a light table. The need is for the ability to drag photos around, compare, resize, and edit their basic attributes.

The prototype of this specific interface is interesting enough that I figure it’s worth sharing.

It’s pretty limited, in that it represents a specific state in a larger application and is very early work. It’s also pretty likely to break. Please use or modify it with that in mind.

Note that this is for Firefox and Safari only. Modifying it to work in IE will be a pain, and is not worth the effort while this is a proof of concept.

A note about browsers

The experience using this is significantly better with Safari. Firefox has two behaviors that limit usefulness; one of which is Mac-specific.

1) The previously mentioned crippling, 3 year old bug for OS X. Therefore, there is noticable chop when dragging.

2) Firefox image scaling is pretty terrible. Here is a zoomed comparison of two scaled images (FF then Safari).

With that said. If you’re using a Mac to view this, please use Safari. If you’re interested, you can download the source and images.

Slashdot has long been a significant source of daily news for the tech community (dupes, typos, trolls and all). Over the past year, however, I’ve noticed a gradual decrease in its importance as a source of information. Others around me have echoed similar thoughts, so I gathered some statistics to see if this is a growing trend.

Without access to long-term Slashdot usage metrics, I looked at the number of comments per story for the last 7 years by writing a few scripts to walk through the Slashdot daily archives, cache them locally, parse the data and then generate a file with the output. Here are the initial results:

The yellow period in the data is due to Slashdot wonkiness. That range of time doesn’t return the usual filtered count (usually the filter is 3) without being logged in, and the overall comment counts are drastically lower than the trend, so I removed it from any conclusions.

There appears to be fairly significant falloff of average comments per story at the beginning of both 2004 and 2005. I also graphed the output of each day’s average comments per post as an xy scatter, reflected below with a trendline.

Note that the count for 9/11/2001 is not drawn (but included in the calculations), as it is exceptionally large.

This second graph reinforces the initial observation that community activity is on the decline.

Why is this happening?

The habits of those around me indicate that the decline in Slashdot activity is due to the following:

• Broad adoption of RSS
  Feeds are ubiquitous, and the result is that everyone can easily customize their daily information exposure — much more precisely than CmdrTaco and the boys can.
• Emergence of social applications
  Yep. Web2.0 apps are killing Slashdot. Relying on the community to process information relevance for individual consumption is a hell of a lot more efficient than Slashdot’s closed door approach (open source indeed). Sites like digg.com and the del.icio.us/popular are far more reliable indicators of what is being talked about.

See Digg Just Might Bury Slashdot at wired.com and The Rise of Digg.com on Slashdot.

What does this mean?

Each of the above linked stories suggest that users are relying on digg for timely information, and go to Slashdot when they want insightful commentary (digg and del.icio.us are for tuning in, Slashdot is for participating). This would imply a drop-off in visitor traffic, but not necessarily community activity (which we obviously see is happening). A sustained decline in both traffic and participation would be bad signs indeed.

If the pool of comments continues to drop, I suspect the signal-to-noise ratio will take an unfavorable turn. If Slashdot can’t do fast, and if it has no community, what remains?

I’ll to continue to track comment activity every few months to see if the trend continues.