Ryan,
My get out is that I don't know much about METS/MODS but I'll try and express why we are not *just* picking them up or any other XML based format. I hope this doesn't come across as a flame - I am just running over old arguments that I probably have said too often. I appreciate you didn't suggest we use METS but I think it needs justifying again.
We could use METS for digital objects, embed MODS for bibliographic stuff and make up our own schemas for each of our domains (entomology, botany, molecular phylogentics, functional ecology, you name it) and we would have integration of data at the application level but not at the semantic level. Effectively each of our domains would have its own XML silo and mixing stuff together would be a complete pain. The attraction of RDF is that it allows the mixing of concepts across domains so we only define things once at a very fine level and can be explicit about what we "mean".
I'll see if I can illustrate this in a naive way by picking one element from the example you give:
<mods:identifier displayLabel="Acquisition number" type="local">27309</mods:identifier>
Do different displayLabel attribute values effect the meaning (i.e. where I put it in my database or calculation) of the value in the element or does the value in the element only mean "mods:identifier" no matter what is in the attribute? So if I put displayLabel="National Insurance Number" or t displayLabel="Barcode" my application may do something different with 27309. How do we do multiple languages for the displayLabel?
The QNAME for mods:identifier from the document would be
http://www.loc.gov/mods/v3identifier
which doesn't resolve. There would normally be a slash or hash on the end of the namespace so that we would get
http://www.loc.gov/mods/v3/ http://www.loc.gov/mods/v3/identifier
but neither of these resolve to anything useful either.
All this may be in MODS documentation but only humans read documentation and then only rarely! Each time we come across a new XML standard some poor human has to go off and read all the PDFs involved before we can get started.
In a sematic web type world all the elements should resolve to their definitions and at that point we can define things like the relationship of this concept to other things and some display labels in different languages etc etc. There is an outside chance that a machine could do something "meaningful" with the information.
Really all XML bought the world is the ability to parse transfer files easily. In the old days when things were space delimited one would have to write a parser to get he documents into memory. Now we can use a generic parser to get them into memory. But XML does not tell us what to do once it is in memory. XML is just a serialization. All the interesting problems are in what is serialized. This is why we lean to RDF/OWL.
I hope this enlightens without putting you off. I expect/hope Bob will have a correction somewhere in what I have said :)
All the best,
Roger
On 8 Nov 2007, at 18:27, Ryan Scherle wrote:
Disclaimer: I don't fully understand all of the issues involved here, as I've only been looking at the biology standards for a few months. I may be misinterpreting some of the points being made. However, I have a good understanding of related standards in the library world, so I hope my comments may be of use.
In my opinion, if you try to put too many external semantics on DC data, you're going to run into many problems in the future, when you interact with groups that have "regular" DC data. It is possible to solve these issues with explicit metadata relationships using existing metadata standards. Here is the metadata for an image object in a repository I built recently:
http://fedora.dlib.indiana.edu:8080/fedora/get/iudl:20008/METADATAAt first glance it looks long and complex, but it's relatively easy to pick apart. The outer layer of metadata follows the METS schema, which is a wrapper format for collecting together different types of metadata.
The first inner object is a MODS record. MODS holds essentially the same type of information as DC, but it allows for more detailed descriptions. This record always describes the artifact in the image, not the image itself. And a big advantage of MODS is that it allows specification of the thumbnail URL that Greg was originally asking about (in the <mods:url access="preview">). Note: It is possible to include a DC representation as well as a MODS representation with a single METS document.
After the MODS record are MIX records containing detailed technical information about each of the image files.
Finally, the mets:fileSec and mets:structMap sections specify relationships between the metadata sections and the actual files. In this case, the hrefs are relative URLs, but they could easily be full URLs or LSIDs.
Now, I'm not advocating that you dump RDF in favor of METS. My main point is that explicitly separating the different types of metadata may be useful. If you would like more information about the specifics, let me know.
--- Ryan Scherle --- Digital Data Repository Architect --- NESCent
On Nov 6, 2007, at 8:48 PM, Ricardo Scachetti Pereira wrote:
Please see my comments in line below.
Bob Morris wrote:
The problem that nobody will take a position on is this:
Is the metadata on an image file, or on an image?
I don't want to dismiss this as a simple problem. We've been trying to knock it down for a long time now. However, I keep wondering why can't we just include information from both (image and image file) in the metadata by using different predicates in each case. See an example below.
Even---or especially if---you stick to DC, you have a problem about what things are part of a description. If the metadata is about the file, then it is reasonable to express, e.g. that it has 1200x800 pixels, encoded as jpeg but perhaps not that it is a a picture of a flea biting a dog. If the image is being described, the reverse might hold.
Couldn't we say the following about an image?
rdf:RDF <tdwg:Image rdf:about="urn:lsid:example.com:image:1234"> dc:titlePicture of my dog Scratchy</dc:title> dc:subjectA picture of a flea biting my dog.</dc:subject> dc:descriptionA description of a flea biting my dog. You get the idea, but an image is worth a thousand words...</ dc:description> dc:identifierurn:lsid:example.com:image:1234</dc:identifier> dc:formatimage/jpeg</dc:format> tdwg:imageDimensions1200x800</tdwg:ImageDimensions> </tdwg:Image> </rdf:RDF>
Even though I bet the RDF isn't valid, I hope you get the point that each predicate refers to either the file or the image, but not both.
If some of these predicates aren't suitable, we can always use some other vocabularies (EXIF?). If you want to refer to what's in the picture, we can somehow point to our familiar biodiversity information objects: taxon name, observation, specimen, etc.
Is there a case where this can't be done?
... rendering clients probably desperately need the pixel size and also information about where to find other sizes of the "same" image.
That's a different problem. We had agreed that LSIDs can't be used if the number of representations of an image is infinite or just very large. Should we be looking at OpenURL or just Web services (and WSDL)?? But that's a little advanced for our simple discussion thread, isn't it?
So, is this a feasible solution, or is there a class of counter examples that I'm missing completely?
Cheers,
Ricardo
tdwg-guid mailing list tdwg-guid@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-guid
tdwg-guid mailing list tdwg-guid@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-guid