Dion Hinchcliffe - Building Successful Next Generation Web 2.0 Apps

Dion Hinchcliffe on Web 2.0

Attending Dion Hinchcliffe's workshop that is covering a lot of ground on the space of web 2.0 as a whole.

Key Points:

His primary message is around value moving up the stack. Value is no longer the hardware or softare, it's the data. Another huge emphasis on 3rd party sourcing. Traditional software's focus primarily on engineering with less of an emphasis on integration is getting inverted. The innovation is in assembly. Have to let 3rd party's source your product/platform and assume more users are using your software from 3rd party locations than from your own. (Twitter has 10x usage through API over web interface.)

Unorganized notes after the jump...

· Click-a-tell.com allows you to send text messages in 700 networks, 200 countries

· 411sync

Core Competencies

· Maintain control over unique, hard-to-recreate information that gets richer the more people use it

· Trust users as co-developers and co-creators

· Harnessing collective intelligence

· Customer self-service

o Using software over the network to service customers

o Turn applications into platforms

o Data-portability is a big problem, dataportability.org, notion of the escape pod

· People don’t want irreversibility

The “Truthiness” Barrier:

· Uncertain ownership and data provenance past this point. Barrier between services and data.

Value has moved up the stack – no longer hardware – no longer in the software – it’s in the data

· Database of intentions – every click, search, tag, recommendation has inherent value.

Cal Henderson of Yahoo: “New features are released to production every 30 minutes”

Innovation in Assembly: Not invented here needs to go out the door. We’re not going to be writing all the software ourselves.

Your data is the fundamental unit of value. Information is the core value, more than software.

· Let your users do it, let your devices do it with sensors, big shift towards sensor driven data

Claims that there are classes of data which are unclaimed citing real estate.

“Heroku” – based on Rails but pushing it further.

Twitter learned a lot and taught us a lot about what didn’t work and scale with Rails.

Think of APIs as a distribution model.

Claim: Cost is in the push services. Adding RSS / 2-way web apps / social net apps / etc is incremental but drives significant value.

Jakob’s law: Have to design your products and services to reflect the fact that most users spend most of their time on other parts of the network.

Think about how to distribute throughout other websites. Think about how to distribute YouTube style through other websites. YouTube is on millions of websites.

Allow plug-in on client sites and allow data to flow

Twitter: 10x more use through API than through Website

CouchDB – very distributed web applications.

Erlang wasn’t being tracked a year ago.

Strategic decision: who maintains your user database? “Log-in with your favorite credentials” – suggests best way is probably doing both.

Google’s social graph API is worth looking into.

“If you aren’t monetizing every page view you’re leaving money on the table” disagree.


3rd Party Sourcing – using infrastructure, data, and functionality, usually live, in a web application integrated from a 3rd party over the Internet.

Disadvantages of 3rd party sourcing: reliability constraint, requires legal skills, have to overcome not-invented here, requires competency in assembly and integration as much as development


Shifting from development first, to integration comes first.

“Web oriented architecture”

Low barrier to distribution model – cut and paste is the lowest.

What isn’t core to our product – get it from the edge of the web.

“You can run your software on the same hardware Google uses.” Key value proposition of Google’s App Engine.

Django is hitting an inflection point with Google stepping up to support the framework.

  • Exposing features
  • Actively supporting developers: documentation and support channels
  • Provide legal, technical, and business reasons to enable: fair licensing, pricing, & support models
  • Unpredictable scaling (Twitter now governs and throttles API usage)
  • Additional effort and expense to maintain each new distribution model
  • Must maintain legal, technical, and community support infrastructure

"The uptake of gadgets has been one of the dramatic stories of the last 2 years." - Really? Are there numbers on this?? I've personally yet to find gadgets useful.

Distribution Models - Not really sure I agree with his clumping of these into a generic 'distribution models' category. Some big differences in intent and capability with each.

  • RSS
  • REST
  • JSON
  • SOAP

Amazon has an API division with it's own Profit/Loss. Demands that much attention and manpower. APIs are first class.

REST and Web-Oriented Architecture
  • "The most commonly used Web service approaches 'in the wild' turn out to be the ones based on the 'grain' of the Web"
  • Litmus test of web 2.0 is 2-way communication - Dion's guess, ATOM takes over RSS in 5 years due to 2-way, RESTful nature

Widgets give users, as opposed to developers, the ability to sprinkle your application across the web. "The more analysis we do - it looks like widgets have even more power overall than APIs." Hmm.

WidgetBox is the largest independent distributor of widgets.


Write-once, run-anywhere social networking apps. Looking promising but still behind Facebook Apps.

"Open Social Spec is amazing. Really intelligently designed."

Over 5,000 Open Social apps. Increasingly looking like "the one". Access to 2 of the 3 largest social networks. Uses ATOM internally. Apache has a module to run OpenSocial. Reach of OpenSocial is currently 50 million.

Key Strategy: Enabling New Consumption Scenarios
  • Anyone: Badges, blogs, wikis, feeds
  • Expert Users: Mash-ups, situational apps, WOA
  • Developers: Web apps, enterprise mashups, composite apps, Office 2.0
"Innovation in assembly."

"JSON hasn't been on the radar for more than a year." That statement made me cringe. Hopefully this was a mental slip :)

"Flash 10 is going to be amazing by all accounts in direct response to Microsoft Silverlight." Note to self: look at what's coming in Flash 10.

For compelling, rich user experiences JQuery is coming out of the gates sprinting hard and taking over. Tiny. Works really well. Sweet spot is with things like JQuery. Open architecture leads to many plug-ins. A big story.

RIA Choices:

AJAX, it's open and organic. Can't do audio and video. A big challenge on the web: open standards haven't addressed rich audio/video media.

Microsoft Silverlight - unclear how popular or how far it will go. (Claim) High-def is better in Silverlight than Flash. Big advantage is that it runs with .Net and comes with Python built into it. Has the most compelling development models of RIAs (agreed).

Sun's JavaFX - Uses Java Applet plug-in. Declarative.

Flex & AIR - Now have great development tools and are becoming compelling.

Dion's rushing and running out of time at this point.