OpenSocial Development Journal – Know Your Friends

At NewsGator, we’re happy to have been selected as one of the early developers in the OpenSocial program. It’s been exciting to build our Didja Hear!? application, and we wanted to share some of our thoughts on social networking and the opportunities, questions and challenges of OpenSocial from the application developer point of view. This is the first in a series of posts that cover the issues we are currently addressing.
 
One of the OpenSocial API’s lets us ask who are the friends of a user. This is a powerful and basic capability of any social networking solution. The primary challenge for us as a developer is understanding what a container site really means when they tell us who are the user’s friends.  
 
Imagine you asked ten random people to name all their friends. Each of them is likely to use different criteria in defining what a friend really is.  Some people might answer the question with a statement like, “Well, I only have two really close friends, but I stay in touch with fifty people.”  
 
Within a single social networking site, the answer is likely to be very consistent. But different sites may choose to answer the question differently. When we were building the Didja Hear application, the rule for Orkut’s answer to the question  “Who are Bob’s friends?”, was to respond with Bob’s friends who already have Didja Hear installed! If we contrast this with Facebook, we can get a list of all of a user’s friends and the information about whether or not they have our application installed.  
 
From the point of view of viral distribution, it’s much harder to prompt Bob to send Sue an invitation to add Didja Hear if we don’t know that Sue even exists.  We’ll have a bit more to say about invitations in upcoming posts…
 
The filter that Orkut was considering around providing friend names is just one example. If we are working with a social network that allows more granular concepts than “friend” or “not a friend”, how will they answer when Didja Hear asks? The most precise answer would be for the site to send us the users along with extended data that indicates how close they are (best friend, friend, business associate, friend of a friend, etc). But if every container site does this, it could be overwhelming to deal with the complexity.
 
OpenSocial is very cool and very powerful. It is already making an impact. But underneath powerful, open concepts there are always important details that need to get ironed out. Over the next several posts, we’ll show you more of the questions we’re considering right now.

-- Brian Kellner is VP Products for NewsGator

Tags: NewsGator, Google, OpenSocial, Didja Hear

Comments

[Kevin Marks]

Brian, it's great to see public discussion of OpenSocial like this. As you say, the choice of who constitutes Friends is up to the Container site, but that is a key point of the API - to make it so that the Gadgets don't need to maintain their own Friends networks, but can use the one that belongs to the social context they are running in.
I think you have slightly misunderstood what Orkut is doing - it is only returning Friends who are able to install the application (ie Friends who are within the sandbox experiment). Clearly you have already been successful at getting people in the sandbox to install Newsgator.
Do bring this discussion to the OpenSocial developer forums too - that should be a place to find more Orkut sandbox Friends

[Brian Kellner]

Thanks Kevin. We primarily wanted to post these thoughts and questions here to get feedback from a broader audience including business stakeholders.

Thanks for the correction on Orkut. The API's have evolved tremendously in the past few weeks and the discussion in the developer forums is like a tidalwave!

Our best products have always come by bringing stakeholder feedback to our developers, so we'll be posting a few more thoughts and questions in the coming days.