I'd been doing some thinking on the relationship between FORMAT and some
of
the issues coming up with TYPE. FORMAT specifies the encoding of an
object, useful
for programs rendering the data stream described by the DC record for
users, content-independent. TYPEs will look a lot like the major MIME
types but instead of providing readable output to a renderer, usually
will
be used to filter or aggregate objects based on a common semantics for
how
the content fits into a larger body of information. FORMAT answers the
question, "how do I decode this object?" and TYPE answers the question
"once decoded, where do I place this object?"
For example, a FORMAT might be image/tiff and the TYPE would be appropriate
to place a scanned page image in the context of an article, monograph,
etc.
That which is needed to describe the relationship of this object to
others near it in semantic space should be
specified in TYPE. So, we would probably need to establish at the top
level of FORMAT name space a set of descriptors that would look a lot like
top-level MIME major types but expressed object semantics instead of
encoding. I'd like to consider just the text or document namespace right
now.
Instead of building an exhaustive list of TYPES, I'd like to push the
complexity of that down one level. How about establishing a set of 2d
level document types as follows:
doc is the top-level TYPE category that describes DLO's. I can think of
four broad classifications of text-based electronic content:
doc.pub document is catalogued as a traditional publication type
doc.msg document is catalogued as a message in an online conversation
doc.file document is catalogued as a computer file
doc.form fill in the blank form
It is possible to have more than one TYPE associated with a DC description
of an object, so none of this is exclusive.
doc.pub.novel
doc.pub.novel.chapter
doc.pub.serial
doc.pub.serial.article
doc.pub.serial.article.section
doc.msg.email
doc.msg.news
doc.msg.ad
doc.msg.press-release
doc.msg.promotion
doc.file
doc.file.homepage
doc.file.README
doc.file.archive
doc.file.db
doc.form.voters-registration
=o=o=
Ideas for the other semantic TYPEs:
two-dimensional visual data:
pic - image/* picture
photo
,painting
, drawing
, graph
, table
,chart
3 or 4 dimensional visual + sound data:
vid - video/* (pic + [sound] + [*]) * time
feature
, short
, ad
, documentary, clip
sound data:
sou - audio/* sound
music
, talk
, ambient
program files:
prog - application/* program
prog.bin
prog.src
miscellaneous types:
dat - data/* data
env - application/? simulated environment
con - multipart/* object container (xml <EMBED>-like aggregation,
CORBA,OLE)
-marc
|