Developer Quickstart

• Every runnable thing within the locker is it's own web service (written in node, python, and ruby to start)
• There is a core web service that coordinates stuff and handles startup, scheduling, etc
• The types of services are generally: connectors, collections, and apps
• Any service once it's "installed" is assigned it's own local directory for storing anything
• Connectors are probably the best way to add value fast to the platform, the more places we enable getting data from the better

Connectors

• Fundamental goal: mirror/sync data from somewhere else
• Store that data in JSON in as identical-to-the-source structure as possible
• Structure the data in a simple archival way, take the long-view
• Serve it back up via as simple of a RESTful API as possible, just make it accessible in bulk upon request
• Be OK with being turned off anytime (graceful start)
• Identify your data to the locker via simple service types (below)
• Generate event notifications when data changes

Collections

• General common data-types, one Collection per type (places, contacts, links, messages, music, photos, etc)
• Need to know how to speak service-types specific to the diverse set of connectors, understand the raw data formats
• Embody the intelligence to merge and dedup from many sources, and handle local meta-data around each datum.
• Register for event notifications to stay up to date efficiently
• Many will consume-from / feed-to other Collections
• Preserve and convey the context of where the data came from ("via")
• Be able to cold start from scratch by "crawling" existing data as well as subsequent updating from pushed events

Apps

• Can talk to Connectors or Collections, whatever data they need
• For now, primary interface is just via web browser
• Should contain all resources needed locally
• Do something awesome for the person with their data :)

Service Types

A "service types" system, mimic mime types for collections/connectors, connectors list what they require and produce, apps can list what types they need to work

A connector advertises which type it produces (page/safari-bookmark), and that type has a spec of common attributes expected within the data for the first part (page), the second part should describe the native format from the source of the data (safari-bookmark)

place/facebook
place/foursquare
photo/flickr
contact/addressbook
page/safari-bookmark
page/safari-history