This is an unsexy directive that could be a profound game-changer for civic society. If you've ever tried to get data from a public entity and had to argue about the particulars of PDF over Excel, this kind of statement is at least a good precedent to work from...some information will still be unparsable, just as much information, period, is still withheld despite transparency laws...but this is a pretty strong statement nonetheless.

Edit: I do think there's a large risk of an unintended chilling effect, as well-intentioned as this initiative. The blog post that was linked to later in this discussion talks about how defaulting to machine data will "ensure privacy"...which can only mean that there will be some process to decide which data should be machine data and which shouldn't...and, I think this will inevitably cause some datasets to just not be released at all, if there are large technical concerns in making sure the data doesn't "violate privacy"

As an example, a couple years ago the Obama administration took down a public database after a doctor complained that a newspaper was able to use it to track evidence against him in a story:

http://www.propublica.org/article/how-complaints-from-a-doct...

The database contained anonymized data about doctor malpractice records...but of course, if you have a doctor with a particularly long history of malpractice in a s single zip code, it's not hard to connect those anonymous records to an identity. So I think there will be some cases where an agency will decide not to put up data because of the fear that computerized analysis of it will reveal things that they don't want revealed.