On Mon, 1 Apr 2002, Houghton,Andrew wrote:
> Has anyone noticed that a number of elements mimic Dublin Core?
> It seems to me given the above proposal and discussion by Geoff
> that LC could just create a Dublin Core profile adding a few
> MODS specific elements to the Dublin Core 15 and then create the
> formal sub-element structure that Geoff describes. That would
> provide the detail LC is looking for while reusing an existing
> Probably not politically correct, for LC's point of view...
Dublin Core has been very effective in allowing for interoperability
between diverse applications, where data can be lumped in big buckets for
various purposes. Particularly unqualified Dublin Core was created for
cross-domain resource discovery. But it doesn't allow for as rich of a
resource description by using the Dublin Core 15 with modifications that
may be needed for some communities.
We originally thought about doing something like a MARC set for Dublin
Core, partially based on Martin Dillon's suggestion in his paper presented
to the LC Bicentennial Conference (he suggested giving people the tools to
create simple records using only the MARC fields that corresponded to
Dublin Core; this involved using MARC systems, so was a little different
than MODS, which uses XML). However, there are several reasons why we
decided to base our set of fields on MARC fields rather than Dublin Core.
1. You say below that LC could "reuse an existing standard" by using
Dublin Core. MARC is an existing standard that has been enormously
successful and suits people's needs for the communities it serves. So we
are already reusing an existing standard.
2. Dublin Core semantics are extremely broad; the advantage of MODS is
that MARC semantics are used, so it is more predictable what type of data
will be in the field.
3. Dublin Core has essentially no content rules. MARC fields do (and let's
not confuse MARC content rules with AACR2 ones).
4. Dublin Core elements do not have anything like subfields/subelements at
this point, so it is completely flat and there's no way to package
together information that all pertain to a particular element. There are
various proposals for what is called "structured values" at this point,
none of which have been entirely accepted by the DCMI community. (see the
work of the Citation Working Group and the proposed "DCMI point" for
5. There is nothing that can be said about DC.Creator/Contributor since no
qualifiers have been approved (e.g. role, affiliation, type of name). This
also relates to point 4 above.
6. We considered using some of the Dublin Core names for elements, but
some did not seem to really describe what the definition suggested.
7. From what we had been hearing in various digital library projects,
Dublin Core descriptions are not rich enough for their needs.
Consequently, either it's being adapted with local qualifiers,
application, etc. in each project or people are inventing their own
metadata schemes. (yes, I know that is what application profiles are for)
8. There is still no agreed-upon XML schema for qualified Dublin Core
(although some work is being done in this area).
As I said, Dublin Core is particularly useful for broad cross-domain
resource discovery. This is not what we are after with MODS, which is a
richer resource description. Interest in MODS we assume is for a more
limited community (although broader than just libraries).