Any time we have a table we should offer export options. We should write an export library that takes already available queries and repackages the result dataset in alternate formats (.csv is really all we need).
If JSON isn't too hard, it's easier to mash up on the client side. XML can suck it.
Steps that are needed to make this happen: 1. Define and document the process in which each new data source implementation requires a CSV, JSON, XML, RSS/Atom, etc. format for that data. 2. Identify the existing data sources that do not have these alternate data formats and open tickets for each data source that requires additional data formats.
We've done some of this (CSVs, RSS feeds). We have now Hive, Pig, and pg access for analysts, and we're going to open our API. Calling this done.