Bob's Lessons

Outline (Bob's Lessons)

Bob's Lessons E. Wilde: Course Wrap-Up

October 13: the composite result of many design choices and tradeoffs between IO-emphasis and IR-emphasis
February 6: a configuration of choices in a complex multidimensional space of organization design decisions
May 3: an intentional arrangement of services for accessing a collection of resources
EMPHASIZE WHAT IT IS, NOT HOW YOU CREATED IT
IO / IR CONVERGENCE is just one of the trends shaping contemporary organizing systems

Bob's Lessons E. Wilde: Course Wrap-Up

The design questions are important whether or not they are dimensions
Some questions are more important than others and must be answered first: the domain, the scope and purpose, the number of users, the expected lifetime
The nature of the items and the purposes of the organizing system are the most important concerns in determining the arrangement of the collection
We can see the categories of organizing systems as design patterns or characteristic functional combinations

Bob's Lessons E. Wilde: Course Wrap-Up

Bob's Lessons E. Wilde: Course Wrap-Up

It was useful to consider collections of physical things, living things, once living things, mixtures of physical and non-physical things because it helped us think more generatively
How would you design a zoo to contain an infinite number of animals?
On the web, zoos and natural history museums are closer together than they are in physical space
On the web, you might start off building a zoo and end up with an aquarium
If the sultan adds more books to his alphabetical camel library what are his design alternatives?

Bob's Lessons E. Wilde: Course Wrap-Up

It is not difficult to expand collection to include digital objects; digital libraries and business applications have done that for a long time
It is more difficult to stretch collection to mean the combination of separate web services or resources to create a composite service or application because this distinguishes the content from the container
- Making the content reusable in more than one container!
- This is at the core of the library's struggle to de-emphasize collecting and emphasize providing access (to resources not in their container)
These collections are organized by document-based workflows in WS-* or by hypertext links in REST architectures

Bob's Lessons E. Wilde: Course Wrap-Up

Architecture is about general principles, top-down, and focused on analysis and modeling
System Architecture is more ad hoc, lower level, and focused on implementation
When Erik and I disagreed it was often not about the issues, but about which of these architectures we were discussing
How the web works and how this particular application works are different levels of description
Link architectures, storage architectures, data/process/presentation integration…

Bob's Lessons E. Wilde: Course Wrap-Up

Community curation (Wikipedia, seed library, crowdsourcing)
Computational / context-aware curation (hidden curation — E. Oreglia)
Curation of the context (in scientific data repositories, need technology and process or the data not reusable)
Quis curatiet ipsos curatos? is the challenge for the semantic web and any organizing system where authority matters because no assertion is inherently the starting point

Bob's Lessons E. Wilde: Course Wrap-Up

Abstract descriptions of services and resources are essential
- In WS-*, abstract interfaces enable transparent substitution
- In REST, resource is an abstraction that hides representations
Unique and persistent identifiers are essential
- Organizing systems that don't have them or don't give access to them are closed and evil!
- Use them to create the Giant Global Graph of Linked Data

Bob's Lessons E. Wilde: Course Wrap-Up

Is a food order a resource? How does this determine how you interact with it? Decouple the person from the order with the buzzer to tell you when food is ready
Can multiple restaurants in a food court share ordering, payment, and delivery resources?

Erik's Lessons E. Wilde: Course Wrap-Up

Designing systems and ecosystems is very different
- you design and then build a system and then you're done
- you design an ecosystem and then hope for systems to start inhabiting it
Designing an ecosystem is a very different activity
- you don't know the systems that will inhabit it
- you may not even be paid to do it
- but it still may decide the fate of what you're doing

Erik's Lessons E. Wilde: Course Wrap-Up

Publishers now have direct access to customers
- publishers need to learn about the role of intermediaries
- libraries need to understand why they are still relevant
Service Systems are a specific case of Organizing Systems
- the internals of the service providers are opaque
- value creation is based entirely on how to combine services
Talking about abstractions is very abstract
- exposing functions has a very long history (OO, RPC, WS-*)
- exposing resources is a little less popular (REST, message bus)
- choosing between abstractions results in very different ecosystems