John Udell recaps a talk at ETech arguing that what the world needs is not standards documents but rather standard implementations. In other words, rather than writing a specification for the Atom protocol, in an ideal world there would be standard, open source implementations of libraries to produce and consume Atom documents.

It’s an interesting idea in theory, but how can we have one true implementation when there are so many languages and platforms out there? If the one true implementation of Atom is written in Python, the Java programmer has to create the implementation by reading the one true implementation’s source rather than reading a standards document.

I do think that standards bodies should write conformance testing tools to release with their standards, so that you can be sure that your implementation of a standard isn’t broken.