On metadata (was: RE: Status code for creating lock-null resource)

> > Of course, this isn't the only way to address the issue --
> > a client could allow properties to be set internal to the
> > document format (like .doc and .pdf files), and then the
> > server could extract the metadata from the document. But,
> > this requires document-specific knowledge, and that implies
> > brittleness over time.
>
> Not good I agree.

Actually, I'm starting to like this solution (grabbing internal metadata
items from a file) more and more.

DAV is nice, since you can add metadata to any resource. But, let's look at
the applications that create those resources. There are two classes of
applications: those that have some notion of metadata, and those that do
not. Those that understand metadata, tend to store it in their native file
format. Storing metadata in a document format has the very nice property
that moving and copying the resource (such as by emailing it) is guaranteed
to preserve the metadata.  If an application adopted a DAV-properties-only
approach, it would lose this characteristic. As a result, there is no
motivation for them to use DAV for properties.

Applications that have no metadata support whatsoever seem unlikely
candidates for adding DAV property support. As a result, we have the current
situation, where authoring clients tend not to make use of DAV properties.

To break the logjam, it seems like several things need to happen:

* DAV clients with good metadata authoring support need to become available.
An ODMA to DAV shim would be very nice, since many word processors have ODMA
support, allowing properties to be set when a resource is created or
updated. DAV needs this capability so that institution-specific metadata can
be set on a resource. This is a fairly fundamental document management
need -- hooks need to be in place so that,if desired, metadata entry can be
made very visible to the content creator.

* Registration of property sets for use with WebDAV needs to happen. A
standard profile for using Dublin Core with DAV is long overdue -- and we're
really close. With standards for formatting and meaning of properties,
metadata authoring clients will have the information they need for their
user interface and protocol layers.

* Existing metadata in documents needs to be exploited. If I've entered
author information into a Word or PDF property field, using the nice
interface provided, I don't want to re-enter that information.

- Jim

Received on Thursday, 21 June 2001 13:11:33 UTC