Here's Mark's example HTTP FIQL (I'm already saying "fickle") query:
http://example.org/feed.rss?title==*great*;ex:rating=gt=4Element names become the query keys. So if your feed language has element <ex:rating>, you can use that as a query key.
FIQL describes rules for constructing URLs based on your content type. Contrast that to HTML forms which prescribe how to construct URLs based on content.
Is that so bad? It's a little too API-like for me. Any Atom feed now has that query namespace imposed on them: If you want to honor queries, these terms become part of your URI space. Amazon, Blogger, you, I all have this URI subspace imposed on us.
If, instead, we define a FIQL content type, application/fiql, and you POST a FIQL document, the server can respond with the query result, or can construct a URI of its own design and redirect to it.
The server retains complete control of its URI namespace.
I'm interested in how we work through this issue because I just suggested a similar strategy for WITSML queries a few days ago.