Lists Home |
Date Index |
Joshua Allen wrote:
>Enriching with metadata I'll buy. You can evolve documents to be a bit
>more machine-friendly, but expecting the documents designed for human
>consumption to be equally machine-processable as a purchase order or
>newsfeed is unrealistic.
Hmm, I agree with you up to a point, but a newsfeed is usually material
intended for human consumption wrapped in a machine-processable envelope
- the same could be said of HTML over HTTP. Similarly the content of
image files is pretty opaque to machine-processing, yet it's possible to
embed useful metadata, as EXIF and XMP demonstrate. Then there is the
use of out-of-line metadata - although the RSS parcel is complete in
itself, there is also usually a link to a permanent version of the
content, about which the metadata also applies. I'll hand-wave towards
WinFS and Apple's Spotlight here as well. Whether the mass distribution
of desktop tools that can make good use of metadata will provide a
positive feedback loop remains to be seen, I think it's pretty likely.
How does this relate to the XHTML and browser issues? Well for a start I
suppose well-formed XML content is easier to wrap up as a payload, and
it's easier to embed metadata in it. Although Jon Udell and others have
demonstrated how XML-based content can add another level of
searchability (through XPath etc), personally I've a feeling this relies
too much on local organisation to be much use on the web as a whole.
Basing content on XML like XHTML may certainly be useful locally, but it
probably is unrealistic to expect it to become the norm on the web.
Perhaps text-oriented content should then just be seen as an opaque blob
that requires a Postelian viewer, whether the material came from a pure
XML doc language, XSLT, +CSS or originated in grandma's text editor.
Metadata can be provided through the delivery mechanism (as in
newsfeeds) or completely out-of-line (as in many RDF-based systems).