Andy Powell wrote:
>
> In describing a Resource-A that has a Creator-B we have two things to
> describe - therefore we should have two spearate, but related,
> descriptions. The '(resource) type' of Resource-A should be part of the
> description of Resource-A. The '(agent) type' of Creator-B should be part
> of the description of Creator-B.
Yes. This is one guise of the 1:1 rule.
> The current discussion seems to be based on the premise that it is somehow
> useful and necessary to indicate the '(agent) type' of Creator-B as part
> of the description of Resource-A. I think this is fundamentally wrong.
I would phrase it differently.
The discussion is based on the premise that it should be possible for
the Agent-type of Creator-B to be available through a description of Resource-A.
In a clean implementation this is likely to be through an indirect route.
In "dirty" implementations, it might be stored embedded in the metadata for Resource-A.
The latter has already happened.
So from here I see just a few alternatives:
(a) consider the Agent-type to be part of the agent-identifier,
according to some not-necessarily-pretty scheme;
(b) have a clean model in mind in the background,
and a clear way to map the dirty metadata into it;
(c) paint ourselves whiter-than-white,
but get left behind by all the implementors.
IMHO a combination of (a) and (b) is both pragmatic and also
does not do any real violence to the clean model.
> I refer to HTML because, looking back on the development of DC, our
> perceived requirement for indicating the type of Creator-B in the
> description of Resource-A comes from the use of META tags like
>
> <meta name="Creator.PersonalName" content="Andy Powell">
>
> by various implementations. Why did people start doing this? Because in
> HTML META tags, there is very little else you can do! It's a kludge. Do
> we really want to base our qualifier principles and data-model on a
> kludge?!
Its a syntactic kludge, but also evidence of a semantic muddle.
I think if you read the proposed definition carefully it not only does not
suggest that the Agent-type is anything like an element-refinement (which is
what the HTML example that you quote implies), but it actually draws attention
to the fact that this Agent-type information belongs with the object, not the
subject or predicate, of the property. I was very careful about that when I
wrote it.
As a matter of interest, I would suggest that the example that you quote here
could be better expressed even in HTML:
<meta name="Creator" scheme="name+type" content="Andy Powell (Agent-type=person)">
where "name+type" is some scheme defined to identify the agent in question with a
string including both the name of it and its type, though clearly it ain't pretty.
HTML isn't really up to it, but harvesting this wouldn't hurt an index too much.
--
Best Simon
|