From my world view, there is a fifth collection type that is capturing our attention: metadata collections. Into this category goes collections of OAI-PMH harvesters (someone else's "digital collection") and selected resource lists (e.g. LII and INFOMINE. This is related to, yet distinct from, licensed collections because there is a) little or no money changing hands and b) librarians have an active role in creating and enhancing records. As with INFOMINE, it operates best when there is a system in in place to automate it. (Perhaps a cross between the Cataloging, ERM, and Digital Asset systems mentioned in your article?)Sure .. the thing about a cake is that you can slice it in many ways. A couple of things ....
In the original post, I was reading the 'collecting areas' off the collections grid. Bought and licensed collections are in the upper left: these are available in many collections and are heavily stewarded. Locally digitized collections are in the bottom left. And managed research and learning collections are in the bottom right. I didn't go on to the fourth quadrant, the open web. This is where some of what Peter talks about belongs: the Internet resource description services. This is an interesting space where clearly there is a lot of library effort, but I am not sure what libraries should be doing individually or collectively there. Indeed, I have spent a lot of my career thinking about this space through various involvements with the UK 'subject gateways' and the Resource Discovery Network, which occupy a similar space to some of the services Peter mentions.
I tend to think of metadata collections in relation to the content collections to which they refer - even if the metadata is managed separately, and aggregated separately. We have always managed collections of metadata - in fact the catalog has been historically at the center of library attention. And those records have been aggregated through many union catalogs, including of course WorldCat. We are beginning to explore how to share metadata for the digitized collections, and have very good technical solutions, but this has not yet matured into a well understood set of services. This is one aspect of what I meant when I said that this activity was in the 'cottage industry' stage.
In general, I agree that libraries will need to manage different flows of metadata more actively.