I tried to read the article twice, but still fail to see what is the 'right way' the author tries to explain. Anybody care to enlighten me on that?
There's also the allusion to custom content-types and versioning schemes and a small bit about using the SF2 web client and crawler for (unit) testing.