discussion of the Creative Commons Metadata work

[Mike Linksvayer <ml AT creativecommons.org>]

Alexander Genaud wrote:
> As per the 'semantic web', I've often wondered if hypertext
> documents wouldn't be better identified and addressed using a
> universally unique ID (UUID/GUID) rather than an URL.

URLs are universally unique identifiers that have the benefit of being 
decentralized.

> It should be noted that there is a massive amount of redundancy
> on the net by related sites as well as search engines and the
> like.   Certainly, this is what Creative Commons is doing with
> related documents.

?

> One could imagine a document referencing sources (or previous
> versions) by an ID. If the referenced material were to change
> location (which occurs often), the reference is lost.
>
>  <a href="http://site.com/12345.html"; uuid"12345.html" />
>
>  &lt;a href="http://site.com/12345.html"; uuid"12345.html" /&gt;>

What resolves the UUID?  You have to trust that the resolver will be 
around longer than the original publisher, and that the resolver can 
deliver the referenced content.  Otherwise you're no better off than 
with using PURL or similar, which can point you to a current URL if 
someone has bothered to keep PURL informed.

OTOH if your identifier is content derived (e.g., content hash) rather 
than just a big random number you don't need a resolver, and maybe there 
will be some means of finding the content (a P2P query now).  It would 
be interesting to see <a>, <img> and the like support multiple URIs, 
some of which could be URNs.  <http://magnet-uri.sf.net> is a workaround 
 that doesn't need additions to XHTML, but does need browser support to 
be useful for normal links and images -- magnet links are currently used 
mostly to bridge the web and P2P.  Another partial solution could 
involve additional HTTP headers.  Perhaps the only practice that would 
cause zero breakage now and hold out some hope of being useful in the 
future would be to publish link metadata, e.g.,

<a href="http://creativecommons.org/";>Creative Commons Rulez</a>

(snippet from elsewhere)

<rdf:Description rdf:about="http://creativecommons.org/";>
<dc:identifier>urn:sha1:7FCKK7IMKPNWUVA2SCJA53R2WHUPNJ5K</dc:identifier>
<dc:identifier>other urn</dc:identifier>
<dc:identifier>yet another urn</dc:identifier>
</rdf:Description>

And see <http://www.w3.org/Provider/Style/URI.html>.

Alexander wrote later:
> Thank you. URN is precisely what I was rambling about. Is it
> generally utalize on the open internet. For example, can I
> search for a document (via google perhaps) by its URN?

Depending what sort of URN you're looking for via Google, you might find 
a Bitzi metadata record (which will likely give you additional content 
identifiers.  AFAIK P2P is the only means of obtaining content via 
certain URNs at this point.

> Perhaps there is an advantage to using an URN in or alongside
> the CC license metatags. Software could easily keep track of
> bibliographic reference for an author and, with the help of a
> good query, do such things as find the root source of specific
> data, or what material references another document.
>
> example URN:
>
> urn:cc:FFFF:md5:5307d294b6ccd9854f2deed8c1628b72
>
> where cc:FFFF might represent an id generated by CC.
>       md5:5307d294b6ccd9854f2deed8c1628b72
>       might be a hash digest of the document.

Recording derivation seems like a problem separable from identifiers, 
and is addressed by dc:source.

--
  Mike Linksvayer
  http://creativecommons.org/learn/aboutus/people#21