I read Reference Lists and Tables of Content and skimmed the referred to Data Packages and Research Object Bundle 1.0 and came quickly away with certain opinion that the days of defining a static data collection specification have pasted. Clearly, where the data for a research experiment can be petabytes in size a static data set is unlikely to be passed around its community of interest. What is needed is a common access protocol for data.
Now there are many hundreds of such protocols in application today. HTTP is an access protocol and one with content negotiation too. The W3C has a boat load of specifications within the SOAP space that are relevant. The semantic web is another deep source. Even the now long in the tooth Object Management Group's CORBA specifications all cover the same ground.
The point of course is that we really don't need another green fields specification and especially one for static data dumps. Instead we need to agree to access protocols and availability contracts. Unfortunately, I have nothing more to add right now as it is not really my problem.