Just to build on Mr. Shawcross's comments, Environment Canada has roughly 40 reusable data sets on a total inventory of 500 potential data sets. Making it available through his programming interface allows them to limit the translation and the metadata to a much smaller subset, which they reuse to all 500.
When we're making data available directly, one data set at a time in a downloadable format, then that particular data set has to have translated metadata. This means that somebody has to take 450 data sets and potentially say that this data set contains this in both languages, the originator is this in both languages, these are the key characteristics of the data in both languages, it will be refreshed on this periodic frequency in both languages, and so on.
While the vast body of data is numeric, which obviously does not require translation, the key preparatory work needs to be translated or it will be just a mishmash of numbers that no one can really use successfully, no matter how creative they are.