>True. How, from the OpenURL, are you going to know that the rft is meant
>to represent a website?
I guess that was part of my question. But no one has suggested defining a new metadata profile for websites (which I probably would avoid tbh). DC doesn't seem to offer a nice way of doing this (that is saying 'this is a website'), although there are perhaps some bits and pieces (format, type) that could be used to give some indication (but I suspect not unambiguously)

>But I still think what you want is simply a purl server. What makes you
>think you want OpenURL in the first place?  But I still don't really
>understand what you're trying to do: "deliver consistency of approach
>across all our references" -- so are you using OpenURL for it's more
>"conventional" use too, but you want to tack on a purl-like
>functionality to the same software that's doing something more like a
>conventional link resolver?  I don't completely understand your use case.

I wouldn't use OpenURL just to get a persistent URL - I'd almost certainly look at PURL for this. But, I want something slightly different. I want our course authors to be able to use whatever URL they know for a resource, but still try to ensure that the link works persistently over time. I don't think it is reasonable for a user to have to know a 'special' URL for a resource - and this approach means establishing a PURL for all resources used in our teaching material whether or not it moves in the future - which is an overhead it would be nice to avoid.

You can hit delete now if you aren't interested, but ...

... perhaps if I just say a little more about the project I'm working on it may clarify...

The project I'm working on is concerned with referencing and citation. We are looking at how references appear in teaching material (esp. online) and how they can be reused by students in their personal environment (in essays, later study, or something else). The references that appear can be to anything - books, chapters, journals, articles, etc. Increasingly of course there are references to web-based materials.

For print material, references generally describe the resource and nothing more, but for digital material references are expected not only to describe the resource, but also state a route of access to the resource. This tends to be a bad idea when (for example) referencing e-journals, as we know the problems that surround this - many different routes of access to the same item. OpenURLs work well in this situation and seem to me like a sensible (and perhaps the only viable) solution. So we can say that for journals/articles it is sensible to ignore any URL supplied as part of the reference, and to form an OpenURL instead. If there is a DOI in the reference (which is increasingly common) then that can be used to form a URL using DOI resolution, but it makes more sense to me to hand this off to another application rather than bake this into the reference - and OpenURL resolvers are reasonably set to do this.

If we look at a website it is pretty difficult to reference it without including the URL - it seems to be the only good way of describing what you are actually talking about (how many people think of websites by 'title', 'author' and 'publisher'?). For me, this leads to an immediate confusion between the description of the resource and the route of access to it. So, to differentiate I'm starting to think of the http URI in a reference like this as a URI, but not necessarily a URL. We then need some mechanism to check, given a URI, what is the URL.

Now I could do this with a script - just pass the URI to a script that checks what URL to use against a list and redirects the user if necessary. On this point Jonathan said "if the usefulness of your technique does NOT count on being inter-operable with existing link resolver infrastructure... PERSONALLY I would be using OpenURL, I don't think it's worth it" - but it struck me that if we were passing a URI to a script, why not pass it in an OpenURL? I could see a number of advantages to this in the local context:

Consistency - references to websites get treated the same as references to journal articles - this means a single approach on the course side, with flexibility
Usage stats - we could collect these whatever, but if we do it via OpenURL we get this in the same place as the stats about usage of other scholarly material and could consider driving personalisation services off the data (like the bX product from Ex Libris)
Appropriate copy problem - for resources we subscribe to with authentication mechanisms there is (I think) an equivalent to the 'appropriate copy' issue as with journal articles - we can push a URI to 'Web of Science' to the correct version of Web of Science via a local authentication method (using ezproxy for us)

The problem with the approach (as Nate and Eric mention) is that any approach that relies on the URI as a identifier (whether using OpenURL or a script) is going to have problems as the same URI could be used to identify different resources over time. I think Eric's suggestion of using additional information to help differentiate is worth looking at, but I suspect that this is going to cause us problems - although I'd say that it is likely to cause us much less work than the alternative, which is allocating every single reference to a web resource used in our course material it's own persistent URL.

The use case we are currently looking at is only with our own (authenticated) learning environment - so these OpenURLs are not going to appear in the wild, so to some extent perhaps it doesn't matter what we do - but it still seems sensible to me to look at what 'good practice' might look like.

I hope this is clear - I'm still struggling with some of this, and sometimes it doesn't make complete sense to me, but that's my best stab at explaining my thinking at the moment. Again, I appreciate the comments. Jonathan said "But you seem to understand what's up". I wish I did! I guess that I'm reasonably confident that the approach I'm describing has some chance of doing the job - whether it is the best approach I'm not so sure about.

Owen


The Open University is incorporated by Royal Charter (RC 000391), an exempt charity in England & Wales and a charity registered in Scotland (SC 038302).