On Wed, 30 Apr 2003, Ben Ryan wrote:

> All, We are currently generating identifiers for objects and metadata
> records within our authoring tool. These are URNs of the form
> urn:hlsi-package-GUID or urn:hlsi-section-GUID or urn:hlsi-object-GUID
> or urn:hlsi-metadata-GUID.
>
> In reference to Andy's points.
>
> 1. We use URN so that they are persistent. I will be shortly registering
> a namespace for the HLSI URNs, though this does not appear to be
> straightforward.

But what do you mean by persistent - if your requirements for persistence
are only over the 10-15 year timeframe, then solutions based on http URIs
are OK (IMHO) - if you want persistence over 50-100 years then solutions
based on http URIs are probably not be OK.

> 2. We use GUIDs to ensure that they will be unique.

Yup

> 3. The identifiers will be resolvable through the HLSI web site. You
> will get a URL that may be used to retrieve a representation of the
> thing the identifier identifies, crystal clear :)

But the thing I end up seeing in my browser is the URL that you generate -
that's what I end up bookmarking, so that is the thing that has to be
persistent (over the time frame that I want it to be persistent for).
Therefore, in practice, it is the URL that is the identifier in the
interaction between your server and my browser.

> 4. Not useful in web browsers, but they are not designed to be. They are identifiers not locators.

I would argue that a locator can function as an identifier (over the time
frame that we are talking about).  (Almost all of the URLs on the UKOLN
Web site that worked 6 years ago, still work now, there may be some staff
pages that no longer work - my gut feeling is that they will all still
work in 10 years time).  Furthermore, I would argue that an identifier
that doesn't also function as a locator isn't as useful as one that does.

> 5. Sounds like a job for code bashers

I'm not sure I understand you here.  The identifier mustn't change, so
there is nothing for code-bashers to do??

> 6. All identifiers are assigned by the software.

In an ideal world yes... But that's a bit like saying that all HTML is
created by HTML editing tools.  In practice that still isn't the case, and
HTML would never have taken off if you had needed a tool to create it in
the early days of the Web.

> 7. We are using the URN system as this allows us to register the
> namespace and then use GUIDs that will guarantee uniqueness as long as
> nobody deliberately mis-uses the HLSI namespace.

Yup.

> 8. The identifiers could be printed, don't know about dictated very
> easily. I do not see how identifiers can be short if they need to be
> globally unique.

Again, all the URLs on the UKOLN and RDN Web sites are globally unique and
(reasonably) short and human-intelligible.  *But*, these kind of URLs fail
requirement 5 (the URL changes if I move the resource from the RDN to the
UKOLN site) therefore they are not good as identifiers for learning
objects.

Andy
--
Distributed Systems, UKOLN, University of Bath, Bath, BA2 7AY, UK
http://www.ukoln.ac.uk/ukoln/staff/a.powell       +44 1225 383933
Resource Discovery Network http://www.rdn.ac.uk/