Re: For Shawn

[cfa@ozemail.com.au (Shawn Callahan)]

Jim

You'll probably hate me saying this but in many ways you have an 'ideal'
situation in terms of metadata.  Firstly, the metadata in your data
warehouse is essential to the success of your organisation.  Without it you
couldn't provide effective products.  I have seen many situations where the
metadata exercise was a seperate entity without any clear link to the
organisation's survival.  In these cases people see metadata development as
the right theng to do but find the directory difficult to populate and even
more difficult (impossible?) to maintain.

Second, and maybe I don't understand your setup entirely (actually I know I
don't), but it appears that a proportion of the metadata can be generated
or extrapolated.  This certainly cuts down the amount of work required.
Jim, how do get the orginal metadata into the system?  And do you (your
organisation) create the datasets or are they sourced elsewhere?

I have always viewed metadata serving three purposes: 1) to find datasets;
2) to manage datasets; and 3) to interpret datasets.  Finding and managing
datasets require relatively few attributes (say 10) especially if they
include author name, author organisation from which an extraordinary amount
of additional information is inferred.  What I mean here is that if I see
the author is Jim and I know Jim is renowned for his rigorous development
of geological  datasets I will infer that the dataset in question is of
relevant quality.  I'm sure this is how we determine which datasets could
be useful.  Much more metadata is required, however, when I attempt to use
that dataset.  This is where Attribute Accuracy, Spatial Accuracy etc come
into play.

Your thoughts of these ideas would be appreciated.  I'm sure there are
plenty of holes.


Shawn Callahan
Strategic Management Sciences