I was at the Gov 2.0 conference in Canberra earlier in the week and found that compared to the talk around social engagement through Twitter and Facebook, the whole concept of open data and APIs took a back seat for much of the event. APIs were mentioned by speakers, but I did not get any sense that the majority of the attendees were thinking about APIs and mash-up-ability of data as much as I do. I also wasn’t sure that everyone knew what an API was, or why you would want one.
So we asked our Director of Product Planning, Don McIntosh to write an article about what APIs are, and why they’re important. This is what he has to say about APIs.
With social applications, there is a clear and obvious use that everyone can understand, and the staggering traffic volumes for these sites make the topic all the more compelling. But what about open data and APIs? Why should we pay them any attention and how do we benefit from them?
An API is an Application Programming Interface. Web based APIs, sometimes referred to as Web services, are growing at a phenomenal rate. Basically, instead of information being presented in a predetermined manner through Web pages, APIs allow other applications (iPhone apps, Websites, MS Windows applications….) to extract specific chunks of information and combine it with other information in all kinds of ways to serve a specific purpose. Jim Ericson from Information Management blogged about this, and he included a good description of how Web services get used:
“Now think of all the thousands of iPhone apps and how they amalgamate all kinds of Web services. You open your commuter traffic app, it calls on traffic information services, Google maps, a weather forecast and maybe an ad for public transportation. One browser app, many (API) calls.”
Jim also mentioned how prominent APIs are becoming. For many popular websites, the network traffic generated by APIs actually exceeds the direct Web traffic. And that’s expected to continue. Perhaps even more interesting is the fact that these days, you don’t even need to be a programmer to use Web APIs. If you have played with Yahoo Pipes, or similar mashup tools, you know what I mean. Basically, these tools are empowering end users to create their own custom applications. Just drag and drop – no coding required.
So, they’re useful, widely used, accessible even to non-programming types, and becoming more popular by the day but what in particular makes them so important in a Gov 2.0 context? I’d summarise it by saying that it’s about making it possible (and easy) for those outside of government to present statistics in a context that is meaningful and useful for them, and that can help facilitate informed discussion and decision making. If I want to provide a service to help people decide where to live, I could combine census statistics such as occupation, income, and age and mash it up with information about the location of shopping centres, pubs etc from a different service. I could achieve the same by gathering all the data into a database and building my service on top, but by accessing the data through an API, my information can remain current, and my queries can be run by calls to the API, saving me from the complexities and resources required to process the data myself. I can also leverage other services such as Google maps to present results. And of course, thanks to mashup platforms, this kind of application might just be something that an (non-programmer) individual does to satisfy their own interest. Either way, it makes it much more possible for people to take government information and use it in ways that government may never have chosen to do.
From a data provider’s perspective, there are many things to consider when looking at providing APIs for direct data access and querying.
1. API vs other means
An API can facilitate innovation, and help automate services that other organizations may provide based on the data. It can also provide transparency by not colouring the data in any particular way, but leaving it open to others to render analysis of the data in their own way. On the other hand, if representing the data in certain ways is useful in promoting an organization’s mission, then it might be best to concentrate on delivering the appropriate views and/or viewing tools for the data. Or in some cases, it might make sense to do both.
2. Risk of abuse
Gartner analyst Andrea diMaio noted that separating data from its source and having no clear way to let consumers understand its lineage or quality runs a great risk of it being misused, or deliberately doctored to represent the “facts” that best suit the application builder. What does this mean to the organization providing the data? Providers of official statistics go to great lengths to defend against this possibility yet by providing data through APIs, they may in some way increase the risk of this happening. Perhaps one way to look at it is to realise that this can happen anyway, without APIs. And it is probably unreasonable to expect a provider to do more than provide accurate quality information alongside their data (and even make it queryable through the API) so that users can make informed choices about what constitutes valid use of the data.
3. Data privacy protection
Many statistical agencies have “remote access data laboratory” services to give researchers the ability to perform detailed analyses on their data. There are typically manual checking processes in this, to ensure that researchers’ queries do not breach data privacy laws by identifying individuals from the data (something that is very easy to do, even when data has been anonymized). A provider would need to determine what privacy risks are posed by making the data available through an API, and ensure that appropriate safeguards are put in place.
4. Resources
An API call results in some amount of processing. Depending on the specifics, such as the type of query and the volume of data, the level of computing resources required can be quite significant. In the beginning, one option may be to limit API use to a few specific applications, and expand that over time. Alternatively, the API could impose certain limits for any single user. This is the approach that Twitter uses to manage the enormous demand it generates.
