(4) From Growth to Explosion

• The Internet started growing in the late 1970s [http://www.internetsociety.org/internet/what-internet/history-internet/brief-history-internet]
  • TCP/IP established robust and reliable global connectivity
  • Successful applications were email and file transfer
  • Internet as connection infrastructure and not information infrastructure
• The pre-Web Internet was always point-to-point
• Applications had specialized and specific interfaces
• 1989, the Web dissolved application boundaries
  • Different services/applications could be interlinked
  • The same client could be used to interact with all of them

(5) Historical Internet Traffic (1993-2004)

(6) Simplicity as the Price for Unification

• Network effect trumps polished verticals
  • Web pages were rather limited for a long time
  • Interlinking and ubiquitous clients compensated for that (Service Fusion)
• IoT today is mostly at the pre-Web Internet stage
  • Well-defined and limited application domains
  • Specialized services that can only be consumed by specialized clients
• Today's apps are the ghosts of pre-Web client/server approaches
  • Tailored/convenient to do one thing in one silo
  • App management is a well-known problem for providers/consumers

(8) Invisibility

• No Web-level identification of resources/services
  • URIs often are resolvable HTTP URIs but don't have to be
  • Identification (even without interaction) creates a lot of value
  • Identity and identification are hard problems for real-world resources
• Data Fusion works best with known and managed identifiers
• Robust naming schemes are needed for all WoT scenarios
  • The network effect needs relatable identifiers
  • Actionable identifiers are a (big) plus but not strictly required

(9) Discoverability

• No easy way to find existing (i.e., identified) resources
  • Will there be a Google for the IoT? How would it work in today's/tomorrow's WoT?
  • Something similar to schema.org [http://schema.org/] may be necessary/helpful
• Discovery in IoT probably also needs a robust update channel
  • The set of discovered resources can change a lot over time
  • Subscribing to change messages allows clients to stay updated

(10) Siloization

• Many IoT scenarios are tightly controlled verticals
  • Easier to design and implement for vendors (engineering legacy)
  • Easier to control and monetize (walled gardens are profitable)
• A lot of value can be created by working across silo boundaries
  • Avoid warehouse approaches selling centralized integration
  • Prefer ecosystem approaches that with a decentralized architecture

(11) Precious Snowflakes (aka NIH)

• Starting from scratch often is easier than reuse
  • Many existing designs are not well-documented and extensible
  • In many cases there is no incentive to reuse
• Defining domain models can be hard work
  • W3C should stay away from specific domains and focus on the connective fabric
  • Domain models should be left to domain experts and their organizations

(13) What is the Web?

• Three pillars are required for something to be truly webby
  • Global Identification
  • Uniform and Stateless Interaction
  • Self-describing and linked representations
• W3C is trying to establish best practices
  • Web of Things Interest Group (WoT IG) [http://www.w3.org/WoT/IG/]
  • Data on the Web Best Practices [http://www.w3.org/TR/dwbp/]
  • Spatial Data on the Web Use Cases & Requirements [http://www.w3.org/TR/sdw-ucr/]
• Lots of work to be done and more Web expertise required
  • Finding the sweet spot for value creation

(14) What's Different About Web/WoT?

• Data Science often assumes centralized data management
• The Web is not only distributed, it is decentralized
• Embrace the Big Data Vs:
  • Volume: Working on a dataset not always (easily) possible
  • Velocity: A robust model for how to do (incremental) updates
  • Variety: Handling data in a variety of models/metamodels
  • Variability: Variations in meaning may need contextual analysis
• The Web focuses on interaction/service models, not data models
  • Describe as little as possible (only the service surface)

(15) Industry vs. Web

• Industry wants to package and sell
  • One integrated and elegant solution for one scenario and/or vertical
  • Also allows relatively easy client-side toolkits/libraries
• Users want to ignore provider-specific things as much as possible
  • IoT has a lot of works best/only in IE feel to it these days
  • Loose coupling means some client-side work is required

(17) Web Data

• Any IoT resources/services should be webby by design
• Web Data [https://github.com/dret/webdata] follows five simple principles
  1. ★: Linkable (use URI for identification)
  2. ★★: Parseable (use existing metamodels if possible)
  3. ★★★: Understandable (document the model/vocabulary)
  4. ★★★★: Linked (relate to other resources by URI)
  5. ★★★★★: Usable (communicate how your resources can be used)
• Web Data is REST without the service parts: data on the Web should be hypermedia
  • REST calls this HATEOAS and adds behavioral constraints

(18) A WoT Service Model

• Activity Streams (AS) [http://activitystrea.ms/] were invented for social services
  • A model for social activities/actors across various social platforms
  • W3C Social Web WG [http://www.w3.org/Social/WG] has been chartered to standardize AS2
• ASbase [http://dret.net/netdret/publications#wil15c] uses AS as the connective fabric for IoT things
  • AS has a (social) base vocabulary that can be used for basic functions (such as following)
  • AS is extensible with new activities (their identifiers and data models)
• ASDL [https://github.com/dret/ASDL] can be used to document extensions in a structured way
  • ASDL is not intended to formally describe AS extensions
  • It is similar in scope to Sedola [https://github.com/dret/sedola] by focusing on making documentation discoverable

(20) Conclusions

• IoT value is multiplied by thinking beyond single IoT islands
• WoT can be the SOA solution for interconnecting IoT islands
• WoT is in its infancy and needs experimentation and experience
• Industry wants to sell IoT but has little interest/expertise in WoT

(21) Thanks! Q&A