W3C home > Mailing lists > Public > public-xg-lld@w3.org > January 2011

Re: vocabs, metadata set, datasets

From: Karen Coyle <kcoyle@kcoyle.net>
Date: Fri, 21 Jan 2011 12:03:20 -0800
Message-ID: <20110121120320.60645xvwknxq2xzc@kcoyle.net>
To: Thomas Baker <tbaker@tbaker.de>
Cc: "gordon@gordondunsire.com" <gordon@gordondunsire.com>, public-xg-lld@w3.org
Quoting Thomas Baker <tbaker@tbaker.de>:

> If four records were grouped into a serialized blob, given
> an identifier, and managed a particular database as a whole,
> would that blob also be considered a record?  I'm not looking
> for an answer, just asking the question...

And a good question it is. This comes back to our definition of  
"record" which is what we are struggling with. I think that "record"  
can be anything you define it as, and will be different in different  
situations. For example, METS is a kind of a wrapper for complex data  
that can contain things like a bibliographic record among other data  
(administrative, etc.). Probably you could call a METS instance a  

What I think is difficult in library land is that we've had one  
definition of record for a very long time, and it's hard to get people  
to think differently. So that becomes one of the issues for the LLD  
group -- working on how we make facilitate that transition in thinking.

> I'm willing to believe that most records _do_ have a single
> focus, but administrative data is a good example.  I took
> a few minutes to look up some examples of library records,
> and the first one I saw had information along the lines of:
>      Berlin: Springer Verlag, 1992.
> ...which I would be more inclined to translate into triples as:
>     :X dct:date             "1992"
>     :X dct:publisher        :Y
>     :Y ex:name              "Springer Verlag"
>     :Y ex:location          "Berlin"
> ...rather than as, say:
>     :X dct:date             "1992"
>     :X dct:publisher        "Springer Verlag"
>     :X ex:publisherlocation "Berlin"
> ...where "Berlin" is directly an attribute of resource "W" --
> which, among other things, would lose the relationship between
> "Berlin" and "Springer Verlag".

In fact, library data does a terrible job with publishers in general.  
The publisher datum is not a reference to a corporate entity that is a  
publisher, it is a transcription of the name of the publisher from the  
title page and in no way is intended to link to an entity. Ditto the  
place. That's in essence a quote from the title page. Oftentimes (and  
I just had a discussion about this on some list and I am totally  
losing track anymore) what is on the title page is not the publisher  
but the imprint ("Penguin Classics"). In that case, that is what you  
will find in the bibliographic record. This datum should not be called  
"Publisher" at all, but that's the shorthand that is used.


illustrates some of the difficulty of deciding what is a publisher.

But in any case, the actual message is:

This book has "Springer Verlag" on its title page.
This book has "1992" on its title page
This book has "Berlin" as the place of publication on its title page.

And if you suggest, as I did on some list, that it would be a good  
idea to have a reference to the Publisher of the book, as an entity,  
you will face wrath, as I did.


> Tom
> [1] http://www.dublincore.org/documents/abstract-model/#sect-3
> --
> Tom Baker <tbaker@tbaker.de>

Karen Coyle
kcoyle@kcoyle.net http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet
Received on Friday, 21 January 2011 20:03:57 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:35:56 UTC