EML

SBC-LTER has been investigating using EML and Metacat for our site's
bibliography. In creating and displaying my xml, I've come up with a list of
potential changes to the eml-literature schema, which dont affect the validation
of current citation eml docs. At the network office, James Brunt and Mark
Servilla are also pursuing something similar while constructing the network's
litdb, and will have some input soon.

My goal was to create a hanging-paragraph style display of our site's
bibliography, with links to the paper and to the dataset as appropriate. EML's
imports/includes were a bit intimidating, so instead, I chose to start from
scratch and write a dtd that looks like eml-literature, but included just the
basic tags needed for the typical hanging-paragraph bibliography display. So the
dtd is kind of an "eml-lite". But in the long run of course, my wimpy dtd should
disappear.

What I've done so far: Create a test xml doc of several pubs with this dtd. The
citations are mostly fictional to suit my stylesheet testing needs. I created a
stylesheet to display it in a typical hanging-paragraph list, filtered by type,
then sorted by year and author. Looking for some thrills, I went ahead and
inserted it into metacat, and mapped the stylesheet (with help from Sid). This
seemed to be a good way to show what I had in mind (not to mention, see
if it really worked). Since most of the citations were created for testing the
stylesheet, most of the url links really dont go anywhere. It was the general
format that I was interested in. You can see the results at:
http://sbc.lternet.edu/catalog/metacat?action=read&qformat=sbclter&docid=sbc_pubs_test.1.1
The xml file and stylesheet can be found at:
http://sbc.lternet.edu/external/EML/SBC_publications/sbcCitationList.xml
http://sbc.lternet.edu/external/EML/SBC_publications/sbcCitationList.xsl

What I haven't done: some cleanup, create css parameters, use id refs, searches,
link the title to the existing table-style eml-literature.xsl. I've started on a
script to convert SBC's own html list of pubs to xml.

Here is the list of differences between my dtd and the eml-literature schema:
1. this dtd has a <citationList> as the root element, (eml: no such tag)
2. <citation> is child of citationList, 1 to many allowed. (eml: one only, root
element)
OK, these 2 do affect current docs, unless they are put into another module
(eml-publications?) that uses eml-literature. Having all the citations in one
list is much easier to maintain than the current scheme of 1-citation-per-doc.
Also, a research site is likely to have many repeated authors, and if the pubs
are in one list, authors can be maintained in the additionalMetadata, and
linked with ids.

3. <title> is allowed to have children, mainly so species binomials can be
italicized. Changing <title> from type:string to type:text would take care of
this (?without affecting current string content?)

4. journal, volume, pageRange are 0 or 1, to accomodate in_press/submitted
papers. <pubDate> is already optional. My stylesheet used the absence of a
pubDate to filter out the in-press pubs. Citations may spend only a short time
in this state, but it's very important to scientists to make their newest papers
available quickly.

5. added an optional <contact> tree, since the first author is not always the
person to contact for reprints. The stylesheet looks here first, then at the
list of creators.

6. added an optional <datasetId> so an accompanying archived data package can be
recorded. This is debatable. I was looking for a way to link archived datasets
to the citation, since some journals are requesting that data be published along
with papers. It seemed a better idea to start from the citation and link back to
the dataset(s), rather than including the finished paper's citation with the
dataset metadata, since after a dataset is revised, the paper may belong only
with an earlier version. Also, papers are likely to use data from multiple
datasets. I was partial to the datasetId tag because the rest of the url could
be created with stylesheet variables. However, this method doesnt allow urls for
any other data catalogs to be included (unless you made additional variables).
Chris suggested an alternative - to use a <distribution type="information"> tree
for the dataset. I'm not sure that this is specific enough.

7. added an optional <description> to distribution/online. Actually, I've wished
for this (or something like it) in eml-dataset, too. It provides a place to put
some text which can appear inside the anchor tags in the html. Which would
really help if the dataset link was put here, to avoid having to diplay the ugly
url and instead describe where the link actually goes. But maybe there's already
a mechanism for this that I've missed.

Thanks-
Margaret (sbc-lter IM)