An integrated Atlassian thanks to OpenSocial (Q&A)
OpenSocial started out as a way to blend consumer Web application data over the Internet, but it also has use behind the firewall, as Atlassian found.
Atlassian is one of those curiosities within the open-source world: like Apple, Atlassian doesn't tend to release its software as open source. But as with Apple, the open-source world loves to use its software.
From JIRA to Confluence and just about everything in between, Atlassian's software is broadly deployed within open source. Intriguingly, Atlassian turned back to that open-source community to integrate its own applications using OpenSocial, as I learned in an interview with Jay Simons, Atlassian's vice president of marketing.
Many people tend to think of OpenSocial as a way for Web sites like LinkedIn to share data with the Web, but Atlassian chose to use it to unify its applications behind the firewall. Why?
Simons: Roughly a year ago, Atlassian took a hard look at improving the integration between our eight products and upgrading the dashboard implementation of Jira, our popular issue tracker. Jira had a decent dashboard, and portlets based on our own technology that customers--and often some of our own engineers on different teams--sometimes struggled to use.
We had a choice: stick with what we have (slightly dated, but solid portlet technology), but make it bigger and better (engineers love this one, usually), or go open.
Asking seven independent teams to learn Jira's clunky portlet stack so they could integrate their products with Jira wasn't a popular option internally, and didn't really buy us much, so we intrinsically preferred the open option. There have been a few different standards for portlets over the years--JSR-168 and WSRP the most well known--but they all felt long-in-the-tooth. And then we took a close look at OpenSocial.
Many initially heard about OpenSocial a couple years ago, when it was billed as a Facebook killer--a shot fired by Google, echoed by the dozens of other consumer social networks scrambling to catch up to Facebook.
OpenSocial defines two concepts--an API for defining and working with social data (profiles, attributes, relationships) and specification for gadgets. OpenSocial's fundamental promise was interoperability--write an application once and host it in multiple social networks. Sound familiar? That's what we wanted to do with our own products.
A year later, and we've shipped OpenSocial in the guts of both Jira, our issue tracker, and Confluence, our enterprise wiki, thanks in large part to Apache Shindig, an open-source reference implementation of the OpenSocial container. Both Jira and Confluence are now OpenSocial gadget containers and producers, and the rest of our portfolio produce gadgets these two products can consume.
The benefits to Atlassian should be clear, but the benefits to our customers are also immediate: display build activity from the Bamboo build server, search results from the Confluence wiki, and recent commits to the SVN through Fisheye on a Jira dashboard that organizes that information alongside the issues and tasks related to an individual project. Everything a development team needs on a single page.
So OpenSocial wasn't a way for you to connect Jira with LinkedIn (or kill Facebook, for that matter)? What are the top-three reasons for choosing OpenSocial instead of some alternative integration technology?
Simons: One, it's open. OpenSocial's openness means it is consistently benefiting from the contributions of the community. Dozens of companies now are starting to work with it, from Atlassian to Google, and the spec is evolving quickly. Gadgets are an important ongoing part of Google's strategy across several products, so our customers benefit from Google's investments and innovations, as well as our own.
Two, easier for developers to grok. Writing portlets for Jira was an art form. Gadgets are far more pervasive, and the technology underpinning them--HTML and XML over HTTP--is the linqua franca of the Web. OpenSocial meant our products were immediately more accessible to more developers.
Three, instant interoperability. We not only got interoperability within our own portfolio, but with other OpenSocial gadget containers, like iGoogle and Gmail, and with hundreds of other OpenSocial gadgets, from Box.net to Remember The Milk.
What's next for OpenSocial?
Simons: The OpenSocial community is quite active. The Apache Shindig project is also quite active. The OpenSocial specification 1.0 (it's on v0.9 currently) should be released in January 2010.
What originated as a technology for consumer social networks, is quickly gaining traction amongst enterprise software vendors, and enterprises. We've created a site to explain how we use it, and hopefully we'll see more enterprises use it well beyond the consumer Web for which it was originally envisioned.
The more the merrier.