Am 09.02.2012 16:50, schrieb Huwig,Steve:
> We have a large-scale proof of concept.
> Google indexes the web. It associates metadata with identifiers. It
> doesn't even need the input of the operators of the millions (billions?)
> of machines that it crawls. When someone uses Google, it returns the
> identifier for the item and allows a user to dereference that identifier
> if they so choose.
Yes. And well, if the chosen identifier then actually works.
But what's being suggested or envisioned here for RDA metadata is
something else: To represent all names, subjects, controlled vocabulary
by URIs instead of by text or codes. What does that mean for the
display of metadata? It means that the dereferencing has to take
place /before/ presenting the display or it will contain next to
no human-interpretable stuff. The proof of concept will have to show
if this can be made to work efficiently. And not just for the display
of single entities like book descriptions but also for result lists
as well. Which, in Google's case, consist of text, stored as such
in Google's bowels, readily displayable, not collections of URIs that
need to be resolved first to be displayable. There seems to be at least
an order of magnitude more Web access necessary in our case. Or is the
idea that we store URIs /along with/ the conventional text being
represented by them? If so, I misunderstood the whole concept and need
to apologize. And ask another question.