Linking Data Across Social Networks APIs
Monday Session 5 Space F
Rohit gave a brief description of his company's product: knx.to.
The short description: it uses the APIs of various social networks the user is part of, and aggregates the user's social graph over all of them.
Next a round of introductions.
Some projects represented:
- Dreamboat - Mozilla
- Familysearch - geneology
- Opensocial foundation
- IBM services
- BT - social network aware telephony
- 43things - community site
- is there business value from linking data?
- how? how to make it open and standard so anyone can do it?
- combine graphs of people (the 'address book' metaphor is out of date -- relationships mean more than contact info) # as service providers, how to correlate identities using the partial data available (ie. is "Rohit K" the same as "Rohit Khare")
- is there a policy layer -- "copy rights" for portability, reversibility
The list of APIs (some used by knx.to, some introduced in the session): * facebook * linkedin * twitter
see also the chart that Rohit should be emailing
Issues/problems when integrating with specific APIs:
- privacy settings are not computable (you can't calculate whether you are allowed to share information you got from facebook -- you have to make another API call to ask them)
- you can't get email/phone numbers unless you are special
- it's hard to look up by email
- the data retention policy is unknown -- they can audit you but the policy is not well specified
- heavily throttled
- short lived sessions: 30-60 minutes
- user gets an email notification every time you get a new session
- friends list is first name only
One of the ways to implement the "which of my friends is on this site" is by searching by plaintext email. However, using plaintext email is overkill, since you don't need to reveal that information. Can use hashed email instead.
Webfinger is another way to interoperate between social networks using the email address as the key.
The problems to solve are: