I've been away from this since my flaming away. Just wanted to chime
in on Rob's points ...

On Tue, Nov 30, 2010 at 3:56 AM, Rob Styles <[log in to unmask]> wrote:
> Hi all, my 2 pence worth...
>
> Not a regular here, joining you specifically for the MADS/RDF discussion.
>
> ** Comments so far
> Some of the comments so far are a tad harsh. It's great to see LoC
> doing this stuff even if it's not exactly as one might have approached
> it. They know their data, maybe we should try to be a bit more
> supportive?

Fair enough. But I do think we need to meet each other half way
("their data" is also "our data" in my view), and I think your
comments are helpful (definitely more than mine) towards that end.

> ** Conceptual approach
> I've worked with library data for a long time and it's not simple
> stuff. A common first mistake is often to assume that something like
> the name authority talks about people and organisations when in fact
> it talks about "bibliographic entities" — the names printed in books,
> mostly.

Yes, I get this sort of indirection. But as an author of some of those
bibliographic items, I'm still a person. And there needs to be a way
to bring these two perspectives together. Concretely, if I have a
description of Samuel Clemens in FOAF, I really want to know how to
link that to some description of his pen persona/alter ego Mark Twain.

> These have been modelled and re-modelled over many years and authority
> data has evolved to meet specific needs. It is not an ideal starting
> point for publishing Linked Data.
>
> However, I think authority data could be approached differently to
> MADS/RDF. Where MADS/RDF uses bibliographic terms, many of which come
> from the record structures employed, I would prefer to see real-world
> terminology used. So, a class of "Name" would be a good thing to have,
> then we can talk about names. Where it is possible to identify a real
> person it would be good to use a class of Person (ideally the foaf
> one) and where we know the name is a pseudonym it would be great to
> have a Pseudonym class too. The current MADS/RDF approach remodels the
> authority /record/ where it may be preferable to model the authority
> /data/.

To me, this (natural language terms, rather than jargon) would go a
long way towards resolving some of my impulsive reaction against what
I was seeing.

> The downside to that approach is that it can make round-tripping
> between the syntaxes harder. Consider round-tripping MARC and MARC/XML
> as compared with MARC and Dublin-Core XML?

So this really comes down to what the priorities are for this effort?
Is it absolutely clean round-tripping with legacy data, or is it to
bring library data into the linked data world? Obviously one can try
to do both, but there's some clear tension here.

> I would look again at anywhere you have a structure word such as
> /element/, /list/ or value as they are likely to be describing a
> record rather than describing things from the world.

Right.

I guess in the end, I'd really like the designers behind this effort
to imagine that people other than library people might also want to
use these data in the end, and to imagine how that might work.

Imagine a case where some developer somewhere is writing some simple
PHP application and wants to store some bibliographic data, but also
wants to be able to link into some LoC SPARQL endpoint to enhance it.
How would they do that? How would they know how to get what kinds of
data, to present it how to their users?

Right now, MADS RDF seems to me to be only intelligible to someone
with a library degree, or with an awful lot of free time on their
hands.

And I agree, BTW, with Karen's suggestion that it makes sense to treat
MADS (or insert other library representation) name representations (I
don't, personas?) as distinct from foaf:Agent or foaf:Person, but to
enable them to be linked.

[snip]

Bruce