Apache Thrift vs Protocol Buffers

Although it was initially developed by Facebook, it's now open source under the Apache License.

The IDL for Thrift looks quite similar to JSON / is comfortably human readable.

The current list includes C, C++, C#, Java, Python, PHP, Perl, Ruby and quite a few others.

Considering that Thrift's WSDL documents tend to be smaller and more straightforward means that it has easy to read definition files.

Unlike JSON and XML, because data is exchanged in binary, there is little to no parsing overhead, resulting in significant performance gains.

As this technology's been used behind the scenes by Google for over 10 years, their native libraries can be considered highly robust.

It's free, open source and licensed under BSD-3

A range of addresses can be declared as available without further definition. This allows third parties to use these spare fields for their custom requirements.

Supports multiple programming languages via third party solutions.

If you hope to present your services to other companies this may be a drawback. REST (and "pseudo-rest" variants) has become the de-facto choice, with SOAP a close second. This means anyone connecting to your services would need to familiarise themselves with a new method of communication, and that they are less likely to have an existing toolset for such integrations built into their systems.

The protocol includes versioning, allowing clients on older versions of code to continue to communicate with an upgraded server (i.e. clients and servers don't need to be upgraded in unison).