>Thus;
>
>1971 as t.min
>1997 as t.max
>
>can be relatively easily manipulated by a computer, AND displayed to the
>user as "Date range: 1971-1997" or equivalent.
I don't want to push the Coverage people one or the other, but an
implicit assumption in this statement should be made explicit.
Displaying the range to 'the user as "Date range: 1971-1997" or
equivalent' requires the browser/editor/creator software to understand
Dublin Core semantics. That is, the software needs to know that the
attributes 't.min' and 't.max' should be combined together in a
particular order and the grouping labelled 'Date range:'.
This assumes that the software has been explicitly written to handle
Dublin Core.
An alternative vision for the future (and one I subscribe to :-) is
that much metadata software will be 'generic'. Feed in a configuration
file and it will handle Dublin Core, feed in another and it will handle
ANZLIC or GILS. (I've written exactly such an editor, soon to be
released as a alpha. Some advantages of this approach are given below.)
Generic software is based on a generic metadata model and it is hard
to encode semantic information.
In making your decision about the encoding of coverage, you should
*not* assume that some software will magically massage the metadata to
make it friendly to humans.
If this makes it hard for users to understand relationships between
those individual attributes, then they should be combined in some form.
andrew waugh
PS. The advantage of generic metadata tools for metadata users is
that you don't have to rewrite your software when the metadata
specification changes (e.g. if you add a local extension). The advantage
to the software writers is that you can concentrate on improved software
functionality, not continually rewriting the same functions to handle
new metadata standards (and reach new markets).
The disadvantage of generic metadata tools is that it's hard to encode
the semantics. This largely limits generic metadata tools to those which
interact with humans (e.g. editors, browsers) where you can encode the
semantics, if necessary, as help pages. Or to programs which don't care
about semantics (e.g. search engines). If you need a computer to do
something specific based on the metadata value the program needs to be
specific to the metadata. But, it will be interesting to see if we can
define sets of common semantics...
|