I've lived all my short career in academic libraries thus far on the new-service frontier. In so doing, I've looked around and learned a bit about how academic libraries, research libraries in particular, tend to manage new services. With apologies to all the botanists I am about to offend by massacring their specialty, here is my metaphor for the two main courses of action I see: grafting the new service on like an apple branch to a crab-tree, or hybridizing the new service with existing services, thus changing the library from the ground up.
Each approach works in some situations, it seems to me. Each approach may fail in others. The question relevant to Trogoolies is—how should a data-curation service work? Graft, or hybridize?
Computer systems administration and tech support, for example, are grafts in most libraries I'm aware of. They have their own staffs who do their own thing and don't interact much with other library staff or services except when something breaks. (It could be argued that the introduction of computers into libraries was a hybridization process; this is true, but it doesn't mean that the library organization necessarily hybridized, and in fact I don't think it did.) For the most part, this seems to work fine. Catalogers and instruction librarians don't need to learn how to configure Apache or Tomcat!
MARC cataloguing, oddly, is a grafted service of long standing. I don't wonder some cataloguers feel helpless before the onslaught of outsourcing, stub records, collaborative cataloguing, et cetera. Because cataloguers are grafted onto the library, it's relatively easy to think about their value in isolation from the rest of the library and sort out how to achieve that value without them. There is a roaring segment of the library literature that believes this is both desirable and inevitable; another segment, of course, is pushing back with all its strength. The proof will be in the budgeting.
An example of the graft approach falling down, it seems to me, is the intersection of systems librarianship with public service. Website design and management. Institutional repositories. Even (some) digital libraries. The graft charged with these matters needs either to forge its own public-service links with the patron base, which is a Sisyphean task for a grafted service's typical level of staffing, or it has to go through the tree-trunk to leverage that trunk's existing contacts. Unfortunately, the trunk may or may not decide to be helpful. Cues from library administration matter considerably, I believe; a tree-trunk that is not told to help a graft will simply starve it. Starvation happens a lot, especially when library administration isn't itself clear on the value of the grafted service.
(The other problem with grafted services, of course, is that it's terrifyingly easy to lop a grafted branch off the budget tree. I'm unhappily witnessing that threat to a number of digital libraries and IRs now, as it happens.)
Information-literacy instruction strikes me as a service in the process of hybridization: fully hybridized at some libraries, partially in others. I do know of some libraries that try to treat it as a graft, but that road seems to lead to too much work for too few people, and eventual hybridization to handle the load. Hybridized info-lit programs seem to work reasonably well, though admittedly there are longstanding questions about the general level of pedagogical skill in librarianship.
Collection development seems to be going in the other direction. This was unquestionably a core service not so long ago in librarianship, and many libraries still consider it one. Events have conspired to push it aside, however, from the Big Deal to approval plans to Google Books. What I'm seeing now (cf. 2CUL) is a willingness to confine this labor to an ever-smaller group of people per library, and a growing belief that holding deep disciplinary and existing-collection knowledge locally isn't the crucial asset for collection development that it once was.
(The question of "local" collection development is only just starting to arise. It's an interesting one for Trogoolies! But this post will be long enough as it is, so…)
So let us consider data curation for a moment. Is it so specialized and grant-funding-driven that a grafted service is appropriate? Or should libraries undertake the fearsome organizational work necessary to hybridize it? (Make no mistake, the organizational work is indeed fearsome, not to be lightly undertaken. Instruction did not travel an easy road to its current hybridized state, and this absolutely brilliant preprint (PDF) discusses the horrendous difficulties one research library bravely sought to conquer when it tried to hybridize scholarly communication.)
My cards on the table: I believe that because of the disciplinary knowledge and necessary public-service responsibilities inextricably entwined with data curation, a data-curation service grafted onto the library may succeed in the short run (or perhaps spottily, in one or a few disciplines), but will fail in the long run. To thrive, data curation will have to become part of the library's core, touching—changing—reference librarians, liaisons and "embedded" librarians, selectors, instructors, systems librarians, and others.
We know from multiple research studies on the subject that researchers believe that the sine qua non skill for data curation is disciplinary knowledge. I have had my doubts about that; I still have my doubts about that. But all by itself, the perception is important, because researchers are the gatekeepers for their data and they won't let people they perceive to be disciplinary ignoramuses anywhere near. In practical terms, then, Achaea University's library can dub Ulysses Acqua a data curator, but he'd better not go anywhere near Dr. Helen Troia or her data without either an extensive basketology background (and librarians are often generalists, but nobody knows everything) or the knowledgeable basketology liaison librarian Menelaus Fox to back him up.
Trust me on this one: Menelaus Fox isn't going to move so much as the tip of his little finger for Ulysses Acqua or Dr. Troia's data unless he's told he better. There are exceptions to this rule, of course, and these exceptions are how pioneering data-curation grafts are bootstrapping themselves. If you ever have a chance to hear Marianne Stowell Bracke of Purdue speak, do yourself a favor and go; she is a sterling example of the exception to the rule, and she's what I think libraries wanting to make campus-wide data curation work will need to aspire to in most (if not all) of their discipline-related staff.
In short, libraries considering data-curation programs will almost certainly start them as grafted services; I can't imagine immediate or anticipatory hybridization even being considered. I myself would be very leery of working for any service whose library administration doesn't have hybridization ambitions for it, however. Such services seem liable to wind up in institutional-repository limbo, which helps no one.
When I have time, which I emphatically don't right now, I'm going to reread this book. I remember it having very smart things to say about what I'm terming hybridization and (if memory doesn't fail) the book refers to as "mainstreaming." I encourage librarians to read it with me!