It seems to me that there is a balance to be struck here amongst the cost of persistence and various other kinds of marginal costs. For example, minting an URL is much cheaper (in time and money) than minting, say, an ISBN. There is the (marginal) cost of dereferencing an identifier, which will vary according to the type of identifier and when the dereferencing occurs. There are many other factors as well.

It's also worth noting that the persistence of an identifier is not a binary condition. {grin} There is a spectrum of durability (and cost), from the cheapest and least reliable URL, through (for example) URIs managed by educational consortia or other agencies, all the way up to identifiers managed by international systems.

Perhaps the best strategy here is for Bibframe to avoid over-predicting the solutions, because different applications will require different balances between these costs. For example, describing a book which will last for several human lifetimes may make the marginal cost of minting really persistent identifiers trivial, especially since many different agents will eventual dereference those identifiers. On the other hand, a scientific dataset that is of interest only to a relatively few agents may only "budget" for a local URL in an institutional repository, with the possibility that if that resource becomes more widely-interesting or lasting, the investment in a URI from a regional preservation consortium will be made, and so on.

I'm not sure that the ability for any agent to mint identifiers really represents "chaos". {grin} I think it carries with it the responsibilities to think through the benefits and costs of minting identifiers of various kinds and to make a sensible choice.

---
A. Soroka
The University of Virginia Library

On Aug 5, 2014, at 6:13 AM, Juha Hakala <[log in to unmask]> wrote:

> If URI representations of ISBNs and other traditional identifiers are based on persistent identifiers such as URNs, DOIs, Handles etc., the chances of avoiding chaos are improved because identification of resources (and establishment of links from identifiers to resources themselves) remains a managed process. If any patron can mint URIs to any resource and claim that they are (persistent) identifiers, the whole system is standing on clay feet, starting from the fact that nobody will ever own a domain; we can only rent them.