I agree with Gerard of course, but I still think there is no contradiction between the two things: the minute, careful curation and interface improvement work on Commons; and the mass-preservation, analysis and usage of it as a dataset. They also are interests for different people with different resources, so there is no competition. For instance, I asked some help with WikiTeam's software, and that's in python, while MediaWiki is PHP and JavaScript: by doing both things we are more likely to cover all interests and *reduce* waste of resources. :)
Nemo