[tm-pubsubj-comment] Fw: a good question from Sam Hunting

[Bernard Vatant <bernard.vatant@mondeca.com>]

I had the following exchange with Sam, and I figure I have to feed the list with it.
I think it is a thorny question. Exchange and comments below.

=================================================

[Sam]
I have a question. There seems to be discussion on the PSIs list about
merging based on URIs alone (ie, with no attention to context). To me,
this implies that there is no distinction contemplated between URIs
that are subject constitutors (ie, in topic <resourceRef>s) and URIs
that are subject indicators (ie, in topic <subjectIndicatorRef>s). Am I
missing something?

[Bernard]
Why should they be distinct for? It's up to the PS Indicator to
declare if it represents only itself (Hi folks, I'm the subject)
or something else (Hi folks, the subject is on third floor next
door). Does it change something to the use of the URI for merging of
whatever identifcation purpose?

[Sam]
Let me answer that question by asking another ;-) --

Given a PS indicator PSI (which, as I understand it, is "just" a
URI), is it the case that the PSI always represents (is a proxy
for)  the same subject, when used as the value of an xlink:href attribute
in XTM syntax?

If this is true, does that mean that the distinction between a
subject constituting resource (*is* the subject, "Hi folks, I"m the
subject") vs. subject indicating resource (*indicate*s the subject,
"is on the third floor next door") has been removed from the realm
of XML (XTM) markup, and placed in the domain of the PSI itself?

=================================================

My view of that question is that there are several cases to make distinct.

1. I use a resource that has not been declared as a PSI
as a subject indicating resource (subject indicator) through <subjectIndicatorRef>
I am completely responsible to decide that this resource is a good way to indicate my
subject.
The publisher of the resource is not aware of that.
 I take the risk of seeing that resource disappear without previous notice.
(but the subject still exists, maybe I should indicate it otherwise)

2. I use a resource that has not been declared as a PSI
as a subject constituting resource (subject indicator) through <resourceRef>
I am completely responsible to decide that this resource *is* my subject.
The publisher of the resource is not aware of that.
I take the risk of seeing that resource disappear without previous notice.
(and well, I'm left with a "has-been" subject)

3. I use a resource that has been declared as a PSI,
as a subject indicating resource (subject indicator) through <subjectIndicatorRef>
I transfer to the publisher the responsibility of defining and maintaining stable and
available my subject definition.
The publisher of the resource is aware of that responsibility.

But:
I may turn out that this PSI is auto-referencing ("Hi folks, I'm the subject")
In that case, would it be equivalent to point to it through <resourceRef>?
Looks like it. If there is a difference, it's quite subtle ...

4. I use a resource that has been declared as a PSI,
as a subject constituting resource through <resourceRef>
This one is thorny:
-- If the PSI is auto-referencing, it seems to be a consistent way to do it, similar to
the above particular case.
-- If the PSI is not auto-referencing:
---- either I made a mistake in using <resourceRef>, if I was wanting to indicate the
subject indicated by the PSI, and not the PSI itself.
---- either I did mean it, and my subject is the subject indicator itself (the resource),
and not the subject indicated by this subject indicator. For example I want to assert that
this subject indicator is irrelevant, or obsolete, or I want to say something about its
publisher, whatever.

In a nutshell:
The distinction indicating/constituting still makes sense in the markup, even when using a
PSI.
But the distinction is irrelevant in the case where the PSI is auto-referencing.

I think that should be addressed in the recommendation, and it's the point of ISSUE 1 -
defining auto-referencing PSIs.

Other opinions?

Bernard