this is a mail from last week which went to the wrong list last week :-)
I would like to respond to two things: first the versioning and,
connected to it, the Container usage.
I would like to argue for a versioning for every Schema. If you want to
solve this by just putting it into a PREMIS Container it might not be
enough. If you process parts of the data and proccess them on their own
(like Markus mentioned in his post), the same problem occurs.
One might face the same problem, if in ten years time one wants to
transport old Metadata into the future and (keep the original! :-) and
has to mix them with newer versions of PREMIS.
Besides you are *loosing flexibility* and not getting a real solution
for the versioning problem with the usage of the Container, there is an
advantage. The advantage of container usage I see is, that if you do not
want to change METS (which I am very much in favour of!), that software
might better know what to expect in a Section.
If you do not know the structure of a METS document in advance, it is
alsmost unpossible to process it automatically. That is, because many
things might be embedded. If one uses the PREMIS Container a written
Parser (e.g. with generated Java Bean Classes) could check step by step
what to process.
On the other hand it might be an idea, to change METS in a way, that one
can specify which Entity to expect in a section. There are different
options to do so.
First, create an own PREMIS section (in AMD section) with a specified
attribute (like ="Object").
Or: extend the other sections (techMD, digiprov...) with a specifying
type attribute. Maybe there are ohter solutions, too.
I like the idea having a separate PREMIS section, but I am not sure if
that is a viable way for he METS community?
Van: PREMIS Implementors Group Forum namens Rebecca S. Guenther
Verzonden: ma 27.08.2007 23:04
Aan: [log in to unmask]
Onderwerp: Re: [PIG] review of document for using PREMIS in METS
On Mon, 20 Aug 2007, Priscilla Caplan wrote:
> The DAITSS programming group (Randy Fischer, Franco Lazzarino, Carol
> Chou, and Manny Rodriguez) met on this late last week and these are
> their comments:
> 3. Use of PREMIS container.
> Overall, we preferred to say that if you want to keep all PREMIS
> metadata together you MUST use the container, and if you want to
> distribute it to multiple METS sections then you MUST NOT use the
> container. Then the simple presence of the container would tell you
> that you need not look anywhere else for PREMIS metadata.
> However, Franco pointed out that if you do not use the container, you
> don't get the version (i.e. v1-1 as opposed to v1). So that would argue
> to always use the container. Or change the schema.
To me, it makes a lot of sense to add the version attribute to each
separate schema. It is possible that someone may mix versions-- or that
one schema may not be updated when another one is. What do others think?
I would agree, it is better to specify that if you put all metadata in one
place that you use the container.
> 4. Use with format specific technical metadata
> The proposed rule is, when the same element is defined in both PREMIS
> and a format-specific scheme, always put a value in PREMIS and
> optionally repeat it in the format-specific scheme. The problem with
> redundant metadata is that it introduces the possibility of conflict --
> what if the two values don't agree? So if you allow redundancy, one
> place always has to be declared as authoritative, in this case
> presumably PREMIS. So if only PREMIS is authoritative, that implies you
> can't trust the non-authoritative metadata in the format-specific
> scheme, and the question becomes, are there circumstances it may be
> useful to record non-authoritative metadata?
> Ideally values should be in one place only. We like Marcus Enders'
> suggestion of using XPath expressions.
The reason we proposed keeping it in both places as an option is that
there are programs that stick Jhove output (like in MIX) in a METS techMD
section. It would be easier to keep it than remove it. But we do need to
experiment with using XPath expressions.
> 5. METS structMap v. PREMIS structural relationships
> Agree METS structMap is preferred, but we recommend no redundant PREMIS
> elements for the reasons stated above.
It might be good to experiment with transforming a METS structMap to
PREMIS relationships for the scenario where METS is used as a transfer
format and the repository takes it apart and stores the metadata in a
database as something like PREMIS semantic units-- or like Fedora, where
I'm told that relationships are treated separately.
> 6. Other METS redundancies
> Agree PREMIS is preferred, but we recommend no redundant METS elements
> for the reasons stated above.
> 7. ID/IDREF and PREMIS identifiers
> Franco's comment is that IDREF is global and it might be preferable to
> use KEY and KEYREF. (I'm not sure I understand this so I'll just turn it
> over to you XML experts out there.)
One of our XML experts didn't understand this either, so please explain.