Jill,

Here is a recent presentation I did on getting records with OAI-PMH,
including getting the records one-collection-at-a-time, and then getting
those into spreadsheet form for analysis across a set of records.
https://drive.google.com/file/d/14n6PwKXdhWvbHPn1NcCXAVbg5MX8q4s_/view
Hopefully the link will work.  I can't attach it because it's a big file.
(Obviously, going from XML to spreadsheet will loose some info from XML, so
for a migration, if you start with XML, you would keep working with XML,
and just use spreadsheets to pull status reports along the way.)  It's
something like a 15 minute intro to how OAI-PMH works, and might be good to
share with programmers.  The presentation is geared towards people who
don't do techy backend stuff, and who just want a clearly defined workflow
for getting specific collections' metadata into a spreadsheet so they can
analyze their collection.  The presentation is about tools with a user
interface, and trying to be as accessible as possible.

OAI-PMH is very simple as far as harvesting records.  It was made in 2002
to try and share large record sets on slow internet, and has very few
queries or functionality.  I think that programmers can understand it
pretty quickly.  In short, it sends large sets of XML records.  And you can
query by collection or by date last updated, but not by anything else.  (No
search, which is very antiquated, and programmers from outside libraries
tend to go "naw!" the first time they see the is no search, but really
there is no search.  OAI-PMH envisioned waiting a long time for all the
records and then working them locally, because internet bandwidth was
limited when it was developed.  Almost the only search feature is getting
all records after a specific date, because the internet was so slow that a
major concern was updating harvests without having to rerun a harvest and
wait days for the text files.  The motivations feel strange now.)  Then
after you query, you have the records in XML in whatever format the system
supports.  - You get lots of XML fast.  I think programmers can learn
OAI-PMH quickly, then the next step of working with the nuances of XML and
XML in general or XML from a specific system is what is tricky.

In thinking back, I learned OAI-PMH largely from software testing - testing
open source upgrades to make sure each thing that can be done in OAI-PMH
still worked after upgrades.  And from harvesting projects and setting up
sites with federated search across open collections.  There isn't a lot of
training material for OAI-PMH.  But also, the standard (
https://www.openarchives.org/OAI/openarchivesprotocol.html just the one
page for all the queries) is pretty short.  A way to learn entry level
might be to find the repository URL yourrepositoryurlhere?verb=Identify ,
and then work through the specification and manually make each query.


Stepping back to your bigger goal, you are trying to go from CONTENTdm to
another system.

The OAI-PMH export is just one possible tool.

There are two big limitations for OAI-PMH in migrations off CONTENTdm.

First, your richest metadata used in CONTENTdm probably won't be in the
schema you need or that can be converted for input into the next system.
OAI-PMH will get you the records in XML format and then in various metadata
schema (OAI-PMH carries any records in XML; so what it support is anything
the system can throw at it, as long as that's XML; then the system will do
various schema - native to the system, Dublin Core, MARCXML, etc.).  My
understanding is that CONTENTdm uses qualified Dublin Core.  Which is a
real bummer, because the Library of Congress publishes crosswalks from one
schema to another, and they have MARC, MODS, and Dublin Core all neat and
tidy for you to grab a crosswalk that will go from records in one of those
XML formats to XML in another of those formats.  But they don't have
qualified Dublin Core.  And qualified is way more detailed than plain old
Dublin Core.  So, you can get your records out in qualified Dublin Core
formatted as XML from CONTENTdm.  If you are going to a system that ingests
qualified Dublin Core, then you have the metadata you need.  If not, then
you have to do mapping from qualified Dublin Core to the other schema.
It's going to be labor intensive, and you will likely pull many spreadsheet
reports to anaylze which fields you are using in qualified Dublin Core to
ensure you have all the fields accounted for.  I think if you are going
from qualified Dublin Core to something else on the new system, then you
may as well use the csv exports that are available from inside CONTENTdm.
Either way, you have to map things and make your own crosswalk.

Second, OAI-PMH is about sharing metadata records, not files.  If you have
only single part items, like one sided images, or PDFs, then you probably
will see something in the records that you can use to latch onto and match
to a file.  But, if you have multipart items, like front and back of a
postcard, or a newspaper issue or book where you have an image file for
each page, then that process of matching files to metadata record might be
more labor intensive than for the single part items.  The OAI-PMH may not
give you a list of all the files, and you might get one file, or like you
mentioned, you might get just a container, then you have to use something
else that is not OAI-PMH to get the list of files associated with that
container.

You asked about limiting by collection.  In OAI-PMH, you can do a query to
get sets (usually collections, and in COTENTdm sets will be collections)
using the setSpec verb.  So something like
yourrepositoryurlhere?verb=ListSets which will give you a list of sets.
Then you can get records just for that collection.  So something like
yourrepositoryurlhere?verb=ListRecords&metadataPrefix=oai_dc&setSpec=thecollectionyouwant
. The attached 15 min video has how to do this - to get a list of all the
collections in your digital library, then to get all the records in any
collection.  If you have items that are in more than one collection, then
you can't deduplicate with OAI-PMH.  Instead, you have to figure out a way
to deduplicate.

Meanwhile, CONTENTdm allows csv exports of metadata.  These will have files
listed more systematically than the OAI-PMH records, so for multipart items
you get a better match between the record and the files than you will with
OAI-PMH.  (Except, warning, CONTENTdm sometimes makes a derivative and then
the software only tracks the derivative.  For example, if you uploaded high
resolution images, maybe it made a big but not quite as big derivative, and
then it kept the original image in the stored files, but it did not keep
the name of the original image anywhere in the software.  But, that is a
quirk of CONTENTdm.  I *think* this about tracking derivative files but not
tracking original high resolution files is some versions of CONTENTdm but
not all, so you may or may not see it depending on when each file was
uploaded.)

I think that your best bet is to specifically identify the system you will
go from and to.  So, identify CONTENTdm to what system?  CONTENTdm to what
metadata schemas can your new system ingest?  Then ask whether anyone has a
migration package available.  Because CONTENTdm changed their licensing
model a few years ago to kill off institutionally hosted installs, many had
to move off CONTENTdm, so there are specific software packages available
that work with the CONTENTdm csv exports and the native file structure in
CONTENTdm.

Best,

-Wilhelmina Randtke

On Mon, Sep 13, 2021 at 4:07 AM Jill Ellern <[log in to unmask]> wrote:

> Code4lib folks,
>
> I have perhaps some stupid OAI questions.  We are moving off Contentdm and
> onto a platform with programmers that I’m pretty sure don't know OAI and
> harvesting at all.  We  have been thinking that it would be simple to
> convert our output of metadata that comes in a text format.  However, we
> see now that it drops the set structure (front and back of an image for
> example) especially since we have some collections that have different
> titles for the container (root description) and the images attached.  We do
> see a line with cpd but with different titles, it look like we might have
> to identify sets in Excel.  That sounds like a big job and a pain.  I'm
> thinking there is a better way with OAI but I don't know much about it.
> My thinking is that we can use OAI to move this data instead of text
> files.  I'm sure it has the structure built in...doesn't it?  Is there a
> easy tutorial on OAI?  I’m not finding much for the layperson. And our new
> vendor is pretty new to library land (they are in museum land) and we doubt
> if they know OAI and I don't see easy ways to teach them.  Do you have
> suggestions?
>
> Jill Ellern
>
>