Yes, what Nate said is what I'm trying to say.
Nate, they aren't _trying_ to use the target URL as a uniquely
identifying key. They're _trying_ to use it as... a target URL! They
just can't find anywhere but rft_id to stick a target URL.
But the problem with that is exactly what Nate demonstrated: Based on
existing use and the spec and best practices with URIs, many will assume
that the rft_id IS intended to be an unamiguous identifier, not a target
And there's no way to tell the difference between an rft_id intended to
be the former (as many of us do) and the latter. Another way is needed
to actually supply a target URL in an OpenURL.
You can do it in rft_dat, I guess, in a custom way -- depending on your
context and use case. Sometimes that doesn't work (like when you might
want to add your data to an already existing vendor-specific rft_dat).
Or, you can do what I increasingly do, and just add your own non-openURL
query parameter to a KEV (or your own namespaced XML to an XML, even
easier). &my.url=http://whatever.com or whatever.
You can't count on any existing link resolvers recognizing it, but they
should safely ignore it, and you can't count on any existing link
resolver doing what you want with an rft_id _anyway_.
It remains shocking to me that even DCTerms doesn't supply any way to
provide "end-user access URL" as distinct from "identifier". They are
not always the same thing, much as the semweb crowd would _like_ people
to always make them the same thing, we aren't there yet.
Nate Vack wrote:
> On Mon, Sep 14, 2009 at 8:48 AM, O.Stephens <[log in to unmask]> wrote:
>> What we are considering is the best way to represent a web page (or similar - pdf etc.) in an OpenURL. It looks like we could do something as simple as:
> Wait, are you using the target URL as a uniquely identifying key? URLs
> can change not only location, but meaning. Consider:
> What if www.bbc.co.uk changes address, and www.bbc.co.uk becomes
> something else you want to link to? Then you'd have
> &rft_id=http://www.bbc.co.uk really refer to http://new.example.co.uk,
> and you'd need to use a dummy rft_id to refer to whatever new content
> lives at http://www.bbc.co.uk.
> Unless I'm totally confused as to what you're trying to do, this seems
> like a bad idea. An artificial key is a better choice.
> If you want something like tr.im, use something like tr.im.