Openly Informatics has published a draft proposal for Latent OpenURLs in HTML. The idea is to deal with OpenURLs that crop up in an unauthenticated context, where the server doesn’t know who the user is and therefore doesn’t know what link-resolver to point the OpenURLs at. The solution, building on the ideas of Dan Chudnov and Jeremy Frumkin, is very simple: the link is identified as an OpenURL by including class="Z3988" in the anchor, and the href contains only the query string, starting with the question mark. It would be up to some other service to insert the address of the appropriate link resolver. This would facilitate the development of autodiscovery systems, along the lines that Chudnov and Frumkin have charted. The proposal is to use version 1.0 of the OpenURL standard, which could easily be downgraded to the syntax of version 0.1 by services that needed to.

Openly’s plan differs slightly from what Chudnov and Frumkin had proposed in their paper: Openly embeds the OpenURL fragment in the href of an anchor, where Chudnov and Frumkin had placed it in a div, thereby displaying the query string to the user (if no autodiscovery service intervenes to replace the div with a link dynamically). The only significant difference is how they would degrade for users who don’t have an autodiscovery service: the Openly user would get a normal-looking but useless link, the Chudnov/Frumkin user would get a rather cryptic display of metadata that probably contains no break-points and therefore might stretch the page horizontally beyond the limit of the browser window. I think I would prefer the former, but the work that the autodiscovery service has to do would be trivial in either case, so perhaps we should try both. (Chudnov and Frumkin also propose an RDF-style format as an alternative).

What external services will insert the address of the appropriate resolver? Maybe browser plugins like the one I prototyped or the far more advanced work of Ross Singer, maybe proxy servers like EZProxy, maybe your RSS reader (we could start putting OpenURLs in blog postings!), etc. The important thing, as Chudnov and Frumkin point out, is that it should put the choice of resolver into the hands of the user, not the publisher.