Hi Carl,
On Mon, 14 Aug 2000, Carl Lagoze wrote:
> Dan,
>
> A bit of a clarification here. While the FEDORA page that you refer to
> does indeed indicate 'per-application methods' I would not say that
> FEDORA or we (at Cornell) propose them. The essence of FEDORA is 1) a
> distributed, extensible, and uniquely identified type (behavior) system
> 2) distinguishing between types (behaviors) and their implementation and
> 3) the ability to associate these types (behaviors) with containers of
> digital content. Since our FEDORA implementation is indeed CORBA based
> types are defined as a set of methods and we have constructed samples
> demonstrating extraction of DC metadata using methods such as getDCField
> and getDCRecord. One could also use FEDORA to construct methods such as
> ASK where the argument was some SQLish expression mentioned by you.
Oh, thanks, that clarifies things. There are several layers there and I
had them tangled up: FEDORA as digital-object transport is application
neutral. The set of methods I stumbled across (thanks, Google!) being
just one of several ways we might expose metadata services over FEDORA.
Now that we're starting to see RDF databases with SQL-ish interfaces,
eg. Guha's RDFdb, http://www.xml.com/pub/2000/08/09/rdfdb/index.html
it might be fun to wrap some of these as FEDORA services and see
what it looks like to application developers.
> I agree with your comment that fine method granularity could get out of
> control. However, its advantage is that it offers some rudimentary
> "explain" facility - e.g., one of the 'native' methods in FEDORA is
> allows asking a digital object "how can you behave" in which case the
> architecture returns the list of methods available on the object. We
> recognize that the name of a method is not necessarily self-explanatory
> of its semantics, but its a start.
If there's a URI associated with that name, sounds like a good
start. You could layer on annotations after the fact, eg. distinguish
between methods that have side effects, quality of service metadata (own
disclosure or 3rd party) etc etc. Nice area to research :-)
While this all might sound somewhat esoteric for a dc-general thread,
this stuff is rather timely. Microsoft and others have lately been
investing a lot in Web-based object protocol work, particularly SOAP,
(see http://www.w3.org/TR/SOAP/ ). I've always found CORBA rather
heavyweight; SOAP seems to promise an XML-based "CORBA Lite" for
mainstream Web developers. If this sort of environment for Web
application developers does indeed take off, we'll have to think about
whether it makes sense to define methods/APIs etc for Dublin Core
metadata services...
--dan
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|