Thanks a lot for the comments, Evan.
We've been discussing the compatibility issues for many times within the
PREMIS-EC. The initial conclusion seems to be that maintaining
compatibility with version 1 of the schema is going to be very
difficult, because the revisions of both the data dictionary and the XML
schema have touched the structure of many semantic units that the
existing XML schema compatibility techniques are not going to help. Our
goal then moves to develop an extensible PREMIS schema version 2 and
hopefully provide more compatibilities thereafter. For systems that
already implemented PREMIS v1, we may provide a XSLT file to assist the
We are, indeed, in the process of incorporating some <other>-like
elements in the PREMIS schema v2. I would like to welcome all schema
experts to join us help developing and evaluating the new schema before
it goes public. Anyone interested please contact me at [log in to unmask]
Thanks a lot,
Evan Owens wrote:
> The discussion of version identifiers in PreMIS in METS caught my eye.
> In the XML DTD world, it is really important to indicate the DTD and DTD
> version to which the document instance conforms. We at Portico see lots
> of publisher e-journal XML and SGML files where it is not at all
> clear what DTD goes with what instance.
> In the XML schema world where namespaces are being used to merge several
> schemas, it is not clear to me that anything is necessary beyond
> well-constructed namespaces that indicate proper versioning. Redundant
> information only raises the issue of what to do when there is a conflict
> and doesn't really gain much.
> The larger question for all of us is whether or not we expect to be
> maintaining objects in our archives that have metadata in
> different versions of the same schemas at the same time. I personally
> cannot see any answer to that but yes, but that idea has driven some of
> our programmers crazy. At a minimum, we expect to have technical
> metadata in different versions as it would be painful to have to
> reprocess all objects immediately before implementing a new technical
> metadata standard. And we may well have different METS versions in use,
> at least temporarily. To make that happen, we have to have really good
> control on versioning; in our case, through namespace identifiers.
> There is a cost in maintaining the patchwork set of schemas that we all
> use (METS, technical MD, descriptive MD, PreMIS), each with its own
> versioning and future development path. It may be that at some happy
> point in the future all the standards will be stable and unchanging, but
> I doubt it.
> The interesting technical question is forwards and backwards
> compatibility in standards. There are threads going on right now on the
> xml-dev list on the topic of compatability and extensibility and
> versioning of standards, technically and politically. There is also a
> really interesting white paper on versioning strategies at the W3C site
> that presents a useful vocabulary for understanding this problem:
> Anyone designing schemas should read this and decide which strategy they
> have chosen and ask whether that is really what they want. (grin). We
> are rethinking some of our own internal schemas in light of this paper.
> Evan Owens, Chief Technology Officer
> Portico www.portico.org <http://www.portico.org/>
> [log in to unmask] <mailto:[log in to unmask]>
> (609) 986 2224
> 100 Campus Drive, Suite 100
> Princeton, NJ 08540