[sc34wg3] The Newcomb "howl" was: Re: New TMCL slides

[Patrick Durusau]

Steve,

Apologies in advance for the length but I left all the comments is as 
context may be important in understanding my response.

Steve Newcomb wrote:
> Patrick Durusau wrote:
>> ...It may be true that some roles cannot sensibly appear in multiple 
>> associations, such as Steve's rather clever example of the 
>> Romeo/Juliet and Tristan/Isolde.
>>
>> But, that does not mean that it would be a sensible rule to say that 
>> a role can *never* appear in more than one association.
>>
>> For example, in Louisiana law, not so terribly long ago (I don't know 
>> if this is still valid law), the husband of the mother of a child was 
>> presumed to be the father. Think about that for a moment. Note that 
>> biological fatherhood isn't at issue but legal fatherhood is. This 
>> was not a rebuttable presumption in all but a few cases. But, it was 
>> possible to obtain support from the *actual* father of the child, 
>> without making the child a "bastard" (a legal term in Louisiana law).
>>
>> So, we could have husband role (legal) - mother role - child role, 
>> and father role (biological) - mother role - child role, where the 
>> mother and child roles participate in two different associations.
> Patrick, I think you're choosing to downplay the fact that one 
> association is about biology, while the other is about social 
> convention. They're not the same kind of association, and the 
> mother/child roles are therefore not the same either, even though they 
> have the same names, and some of their aspects are the same.
Ah, ok, what is your basis for the claim "They're not the same kind of 
association...."?

Obviously you must be standing in some perspective in order to make that 
claim. Or as we would say in the TMRM, you are operating with a 
particular legend in mind when you make that claim.

You see I don't really disagree with the notion that *one topic 
represents one subject* but the judgment as to whether that rule is or 
is not being violated, can only occur from the perspective of some 
particular legend.

Particularly if we are going to avoid the underlying fallacy of the 
ontology work that presumes for all of us some particular set of 
ontological assumptions. While I am sure those making such assumptions 
are well intentioned, but, no thanks, I have my own ontological 
assumptions.

 From one perspective (read legend) it may very well be true that using 
your "role playing" analysis (more on that in a moment) that someone can 
say that I am confusing two different types of associations. OK, but if 
from my perspective (read legend) I am only talking about one, which 
perspective or legend gets to control?

You seem to be implying that there is an *uber* legend that encompasses 
all topic maps and that enforces some "objective" legend over of all of 
topic maps. That is precisely the presumption (which I doubt is what you 
mean) that cripples other information technologies in the long run. One 
either has the "canonical" view of a particular paradigm or you have to 
get yourself another paradigm.

Having said all of that, I certainly concede that within any particular 
legend, particularly within particular domains, there may be commonly 
accepted or even standardized identifications of subjects that if you 
are writing a topic map under that legend and in that domain, you need 
to conform to those expectations.
>>
>> I don't think that many people would say that the mother and child 
>> roles in those two associations are different. That is to say, they 
>> meet the test of "one subject per proxy" and yet as roles they appear 
>> in more than one association. (Note that this isn't an analytical 
>> issue. The law of Louisiana defines the subjects in these 
>> associations. That we might model the subjects differently, for what 
>> seem to us to be good reasons, isn't controlling. For topic maps in 
>> legal systems, the subjects as defined by others are what topics 
>> represent.)
> What you seem to be saying is that there is absolutely no difference 
> between social motherhood and biological motherhood. For me, that's 
> obviously not true. And I would argue it's not true under Louisiana's 
> laws, either, as least not as you describe them. The fact that the 
> difference between these roles is obscured by the fact that they have 
> the same name is not relevant, but of course you know that.
>
> If your point is that "the law is an ass" in the precision of its 
> definitions, I'm not sure how that makes your example convincing.
>
> While it may be true that Louisiana's legislators saw no need to 
> explicitly recognize a distinction that had no utility with respect to 
> the purposes they had in mind while legislating, it doesn't follow 
> that no such distinction exists, or that such a distinction should not 
> be made when using a rhetoric that requires one-subject-per-topic 
> (i.e., topic maps).
>
Ouch! OK, how is that distinction obtained if the people writing the 
definition don't recognize it? Distinctions, just like all other 
subjects, can be specified only in the context of a legend.

It may be that you could convince a court that despite legislative 
definition to the contrary that it is useful that have this additional 
distinction, but then you are changing the legend and the subjects it 
defines. Perfectly legitimate to do so but only upon that change does 
the *capability* for the prior association to have represented two 
subjects even come into being. Before that point, it did not represent 
two different subjects. It simply represented one subject, which from 
another point of view was ill-defined.

Unless you want a task like "correcting everyone who is wrong on the 
Internet," I think the perspectives of topic maps authors in terms of 
the subjects they choose to represent is going to have to control.
> If you're arguing that the model of a legal relationship type in a 
> Topic Map must be exactly as ill-defined and ambiguous as the law that 
> it's supposed to represent, I'm sympathetic to that requirement. In 
> fact, I enthusiastically embrace that requirement. But I think that 
> requirement is easily met *without* breaking the one-subject-per-topic 
> rule, at least in the case of your example. In fact, the goal of 
> unambiguously representing ambiguity can *only* be achieved by *not* 
> breaking that rule: by being explicit about the fact that there are 
> two kinds of associations here. And that, in turn, requires that we 
> recognize that the mother and child role types are, in fact, different 
> subjects when they appear in the context of different association 
> types. If we don't choose to recognize that fact, then, as I tried to 
> explain in my howl, our whole notion of what a relationship is lacks 
> integrity and isn't grounded.
>
> What you're saying sounds to me like, "the fundamental nature of the 
> notion of "association" is whatever the author says it is, and that 
> the whole question of what are the semantics of association (roles, 
> the playing of roles, and the unitary nature (the uniqueness) of 
> association instances) in Topic Maps is undefined by the standard or 
> even by convention. I would argue that, at least in my own mind, 
> that's not true, has not been true in the past, and should not start 
> to be true now. I think to claim otherwise is implicitly to claim 
> that, in fact, the "one subject per topic" rule itself is meaningless 
> and without any real constraining effect on topic maps that claim 
> conformance to the standard. I hope that's not your position (but it 
> would lead to a very interesting conversation if it were, so don't 
> hold back).
To the extent that any subject, and I include the notion of association 
as a subject, is represented by a topic map then what other source than 
the author of that topic map would I pick for defining the subjects they 
have chosen to represent?

That does not mean that a particular legend cannot have rules for how it 
defines associations or roles. For example, to add properties to an 
association, the TMDM says I must reify the association and that "For 
example, creating a topic that represents the relationship represented 
by an association is reification." So I wind up with two representatives 
for the same subject. One is the association and the other is the topic 
that reifies it.

Lest that be seen as a criticism of the TMDM, remember as Hoke says in 
"Driving Miss Daisy" "...you was in the car with me when we took the 
wrong turn." And so we all were. I am sure there are a number of things 
that with almost ten years worth of experience and thought about topic 
maps that we could change. That is what standards do, viable ones anyway.

What I don't think the TMDM (or any other legend) can do is define some 
implied notion of subject by which all authors are bound shave. We can 
certainly argue about what distinctions are or are not useful in some 
context but those discussions will always be in some *context.* Never 
outside of a context.
>> My conclusion is that roles *can* appear in more than one association 
>> and still meet the "one subject per proxy" test. That not *all* roles 
>> can do so is interesting but should not lead to the rule Steve 
>> advocates.
> I'm not convinced, not yet anyway. Your example either is not speaking 
> to my point, or I don't understand your example, or there's some 
> subtle but profound difference in our universes of discourse, here, or 
> maybe you just reject my analogy between relationship type and 
> scripted drama.
>
> (If the latter, you must then claim that "role" is a poorly-chosen 
> term of art that misleads people about the semantics of the notion of 
> "association" in topic maps. I can assure you, for whatever the 
> assurance may be worth, that the word "role" was very deliberately 
> chosen with the intent of implying an analogy between scripted drama 
> and association type.)
>
Oh, I don't doubt that many of the terms in the topic maps paradigm were 
chosen for their clarity, immediate and universal apprehension by anyone 
within hearing distance. ;-)

But, your interpretation of role implies more than the analogy between 
scripted drama and association type. You also assume a level/type/method 
of subject identification that leads you to *particular conclusions* 
about how many subjects need to be represented by distinct roles in 
associations.

That is certainly a valid level/type/method of subject identification, 
but if that is the case, then so are all other levels/types/methods of 
subject identification. That within some view there is no distinction 
between biological/legal mother and child, and that may produce what 
others view as multiple subjects or being suboptimal for some purpose, 
but so what?

All that really matters is that the presence or absence of that 
distinction is disclosed so that when I encounter such a map that I can 
take the necessary steps to prepare a mapping to it. It may be like RDF, 
which doesn't make the same distinctions as the TMDM with regard to IRIs 
and so a mapping may be difficult in one direction and lossy in the other.

But that is the nature of making mappings between one choice of subjects 
for representation to another. Within a particular legend and 
particularly within a domain of practice, that will be to some degree 
seamless. Across legends and worldviews, the terrain is a lot more rugged.

[Anyone who has read this far will realize that my position has shifted 
somewhat from my first post where I argued that even within a 
perspective that a topic could represent more than one subject. Steve's 
replies have a way of making me work at being more specific.

The real answer is how many subjects a topic represents is a question of 
where you are standing when you look at it. From some perspective, there 
may be only one. From others, your mileage may vary. The only 
perspective that can control (to me anyway) is that of the author.]
> You'll be missed in Leipzig, Patrick!
>
And I will miss being in Leipzig!

There is a place in the train station (the mall level) that has entire 
roasted chickens. I think the smallest portion you can buy is half of a 
chicken. Highly recommended. On the right hand side if your back is in 
the general direction of the North Church.

Hope everyone has a great time in Leipzig!

Patrick

-- 
Patrick Durusau
patrick at durusau.net
Chair, V1 - US TAG to JTC 1/SC 34
Convener, JTC 1/SC 34/WG 3 (Topic Maps)
Editor, OpenDocument Format TC (OASIS), Project Editor ISO/IEC 26300
Co-Editor, ISO/IEC 13250-1, 13250-5 (Topic Maps)