Hello
Last week I had some other things so I did not manage to do a lot towards the
BibDoc generalisation.
Could You please look at this and tell if it makes sense (from 2 persectives:
1) the consistency of the proposition
2) The compliance with METS standard )
Implementation of this will make Invenio input METS-like, but it will not allow
input from other repositories.
It is very very unlikely that data coming from other repositories will be
formatted according to this schema so maybe it is better not to make it
METS-like
so that nobody gets confused that gerneral METS is supported ?
Do I understand correctly that every METS document is supposed to describe
exactly one object ? This would make the example break conventions of METS.
Here how I imagine mapping of described earleir objects:
- The structMap being a container for objects (every top-level division would
be interpreted as one of entities described in one of previous e-mails)
- Uploading/modifying Bibobjects
A division having a type "BibObject" would represent a single BibObject.
The ID field can be used for encoding versions, permanent ids and
temporary ids (depending on prefixes)
BibObject division should not contains subdivisions, only <fptr> links
to files appearing in BibObject.
example:
<div id="tmp:NewObject1" type="BibObject">
<div TYPE="version" ID="tmp:NewObject1:newVersion">
<fptr ... />
<fptr ... />
</div>
</div>
or new version of an existing record:
<div id="objectId:123" type="BibObject">
<div TYPE="version" ID="objectId:123:newVersion">
<fptr ... />
</div>
</div>
- relations between versions of BibDocs implemented using structLink element
for example:
Linking using temporary identifiers:
<structLink xlink:from="tmp:NewObject1:newVersion" xlink:to="objId:232314:3"
xlink:arcrole="is_extracted_from" DMID="link to metadata in serialised moreinfo
" />
Linking already existing entities:
<structLink xlink:from="objId:2334234:5" xlink:to="objId:232314:6"
xlink:arcrole="is_extracted_from" DMID="link to metadata in serialised moreinfo
" />
Similarly, we could provide relations between particular bibdocfiles.
- storing more-info in Descriptive metadata objects in "binary" format
(serialised JSON) and assigning using DMDID to entities.
This would be simple to implement but clearly would make this part of METS
files non-readable
I think, we should go also for described earleir additional MARC field.
Thanks for comments.
cheers
Piotr
