Does a text citation make a good persistent identifier?

We've been talking about persistent identifiers at RUBRIC: how are people going to name the things in their repository so they can be cited and located in the future? We're looking to the  PILIN project (read a presentation in PDF or text) for some advice. Expect to see more on this topic here over the next few months.

Lots of repositories use Handles to name things and the handle infrastructure for naming things and storing metadata about them is pretty impressive; but others, like USQ have yet to buy-in and some people think that URIs are perfectly good names. Me, I'm still confused.

Anyway, a practical problem: how do you refer to an item in the USQ ePrints repository?

How about a plain-text citation? This identifies something; it's a perfectly good name.

"Sefton, Peter (2006) The integrated content environment. In: AusWeb06: The 12th Australasian World Wide Web Conference, 1-5 July 2006, Noosa, Australia."

My theory was that I could use Google as a resolver to find versions of the paper by doing a phrase search but that doesn't work. The full citation with the quotes yields nothing. Without the quotes Google takes me to the ePrints page. But Yahoo finds it using a phrase search. So does MSN.

USQ ePrints has a version of the paper in PDF format but the actual paper is available from the conference web site in HTML format. As I write this there's no link from ePrints to the version at the conference; I'll see if I can get that added.

You can find the version of the paper at the conference web site by searching  for AusWeb06, then clicking through thedays of the program (I was on Tuesday).

There is a also a paper version of the proceedings – so of course you can find that via  the National Library of Australia. (I searched to ausweb06). That finds the proceedings of the conference: AusWeb06 : Making a Difference with Web Technologies / Treloar, Andrew(ed), Ellis, Allan(ed). Maybe we could update the ePrints repository to use this as part of the citation but if we do that then the links in this post will eventually break – the citation I used  in this post will no longer be there.

At this stage its looking like the plain text citation is not a particularly good persistent name if you want automated resolution, but on the other hand a human being can use it find the paper, and it should work in a post-Internet world.

My understanding of handles is that you should be able to link all the different versions of the paper by storing URLs as metadata against a single name, the handle resolver service could be used to point to any one of those versions, or you could make a service that threw up a page where people could choose.

The handle could, in theory be like this (where n would be a number): 1959.n/"Sefton, Peter (2006) The integrated content environment. In: AusWeb06: The 12th Australasian World Wide Web Conference, 1-5 July 2006, Noosa, Australia."

Except that I have heard some people say that this kind of  local name (that's the bit after the /) is a bad idea and names should be meaningless. I don't understand that bit yet – maybe it's because of the issue I noted above with the wording of the citation.

We're working through lots of practical issues to do with naming, citing and locating things on the RUBRIC and ICE-RSprojects – more soon.