Sean Conman seems to have a lot of good things to say. Recently, he talked about a private markup language he developed, which offers him more flexibility in writing blog entries than ordinary Markdown, RST, AsciiDoc, or other such languages. No one else will likely ever use MOPML (his name for the language), but if it fits his needs, why not?
There's a parallel here between this and the Linguacode data format I developed for IQNavigator/Beeline. No one outside of them will ever use it; our customers will keep feeding us data, it goes through the Boomi translator, and the resulting XML data gets parsed and fed in by the integration code in FrontOffice. I don't even remember all the details of the format, but I don't have to at this point, as I'm not actively working on it.
The professional services people who set up the translation routines in Boomi didn't have to know the whole details, either. The general outline of a Linguacode translator is well-known at this point; they know that certain "magic" values will go in the header stanza, and the data fields from the original lines of text in the input get mapped into specific fields in each data stanza that's generated.
We all have to reinvent a few wheels every so often. And sometimes something domain-specific is more suited to its purpose than anything more standardized could be.