Atom: one format to rule them all?

Recently I’ve had several conversations regarding the Atom Syndication Format. This format is gaining more and more adopters and several big players in the industry are using it. Just to name the big boys, Google AND Microsoft are using it to implement RESTful APIs. When was the last time you heard Google and Microsoft agreed on something? This should hint that Atom is indeed a nice thing happening in our field.

Web services using the Atom format for data exchange encapsulate information using Atom’s standard elements and also define some extension points where needed. Some very useful elements are present in the specification. There are standard ways of publishing individual entries and collections, pagination support, links between resources, among other things useful for RESTful web services.

Depending on your domain model, the amount of data you would put in Atom extension points varies a lot. Some domains such as Google Apps can produce a RESTful model that uses a lot of Atom’s standard elements without needing to define too much extension points. However, if you use Atom to exchange billing information between ISP applications, you’ll probably have to define a lot of extension points. I’m saying this to show that some domains match much better to Atom structures than others.

While talking to Silvano (a very clever working mate of mine) a couple of weeks ago, he asked me if encapsulating everything inside Atom elements was not the same as encapsulating everything inside SOAP. This is a very very good question.

When you’re choosing a format for you data exchange in web services, it’s very important to analyse what you gain and what you lose by picking any given format.

For example, a good rule of thumb about SOAP services is: “WS-* is just overhead unless you have something meaningful in your SOAP Headers” (quoting Sanjiva Weerawarana at ApacheCon 2007).

Atom was designed in a RESTful manner by a very talented group of professionals. Many applications are making use of it to exchange data, and the adoption is growing fast. Could it be a silver bullet then?

That’s where I shall leave my observations. If you consider my example of billing information, you’ll see that most of the data there doesn’t mesh well with Atom’s standard elements. Thus, we’d need to define a lot of extension points, and wouldn’t make much use of Atom’s resources. Putting billing data inside Atom entries would represent an overhead without giving us much in return. In this case, I’d rather use my own XMLs directly over HTTP.

Am I saying that Google and Microsoft made a bad decision choosing Atom? No, absolutely not! Their decision was very good. Microsoft is using Atom for Windows Live API and Google’s using it for most of their applications. What do these have in common? They all manipulate web content. They have a domain where many things are accessible on the web, with lots of URIs, different media types, pagination, categories, tags, etc. Atom makes a lot of sense with web content.

I don’t think Atom is a Silver Bullet for RESTful web services in general. Of course you can choose to always use it, and benefit from the standards and the avaiable tools. But isn’t this true for SOAP as well?

What I do think is that Atom is very close to a Silver Bullet when you’re dealing with web content. Whenever you’re developing web services, choosing the right format for your data exchange is one of the most important decisions. To know well the avaiable options is very helpful, and certainly the Atom format brings a lot to the table when your domain meshes well with it. As long as you don’t think it’s the best choice for every application, go ahead and use it wisely