Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

SUO: Re: Lifecycle Integration Schema




o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o

LIS.  Discussion Note 10

o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o

Focusing on:

LIS: | thing
     | 
     | A <thing> is anything that is or may be
     | thought about or perceived, including
     | material and non-material objects,
     | ideas, and actions. 
     | 
     | Every <thing> is either
     | a <possible_individual>,
     | or an <abstract_object>.
     |
     | NOTE 1.  Every <thing> is identifiable within a system.
     | System identifiers created by other systems and received
     | as part of a data exchange may be stored for future reference
     | as an identification, referring to the originating organisation
     | or system.
     |
     | NOTE 2.  Every example provided for other entity data types
     | declared in this schema is also an example of <thing>.
     | 
     | http://www.tc184-sc4.org/wg3ndocs/wg3n1328/lifecycle_integration_schema/lexical/thing.html

JA: You are saying that you are really only thinking about things
    that have identifiers in given system, or names in a certain
    context of discussion.  This is very significant and needs to
    be elevated to the level of an explicit principle, instead of
    being left implicit in a note that is likely to be neglected,
    in other words, relegated to a hidden axiom or constraint
    whose consequences are not critically reflected on.

MW: I think it is more that we are saying that if you want to
    say something about an object under this standard, you have
    to be prepared to give that object an identifier.  This is a
    practical rather than philosophical statement.  You will find
    elsewhere under representation a quite general approach to
    identification.  Providing this "system unique identifier"
    was more a piece of practical design.  I.e. philosophically
    you can ignore it.

JA: In formal computational terms, a "system unique identifier" (SUI)
    of a denoted object is known as a "gödel number" (gnumb).  There
    are gnumbs for objects in the external world, whether abstract or
    concrete, and there are gnumbs for every finite piece of text in
    the formal language that we use, the latter gnumbs being what we
    are really creating when we "quote" that piece of text.

JA: A measure of how "critically reflective" a formal system can be,
    or help us to be, about these external and internal worlds both,
    is determined by many pickwickian details of the gnumb function
    gnumb : X -> N, where N is the set of nonnegative integers, and
    X is typically thought of as being built up in layers from some
    initial X_0 that we might treat as the "initial external world".

JA: From another angle, if we think of SUI's as "coordinates"
    of objects, that have their meanings relative to a particular
    frame of reference, but may be total gödellygeek from the POV
    of another reference frame, then what we have here is the task
    of intercommunication among codes that we have charged ourselves
    to discharge.

JA: So there's a lot more to say about this,
    but the thrust of it all is that we can
    no longer just shove these issues aside,
    under whatever category label we assign.

MW: Then I suggest you consider this attribute deleted for the 
    purposes of the use of this material within this forum.

JA: I don't understand what you are saying here.

JA: Perhaps you mean to delete Note 1, but that note is
    key to specifying in operational terms what you mean
    by identity, identification, individuals, and also a
    <thing> as "anything that is or may be thought about",
    since those thoughts do not enter into public discourse
    with being translated into signs of some kind.  That is,
    the defining clause for <thing> cannot be separated from
    the issues of Note 1.

MW: What I mean is that there are two separate elements to the data model.
    The first part is a model of "the world".  For this we do not need
    Note 1 or the ID attribute it refers to.

Yes, the world is there.  I will make that hypothesis.
If we knew it directly as it is we would have no need
of approximately iconic models or logical formalisms.
But we got kicked out of that garden a long time ago.
So now it is only in the medium of concepts, models,
signs, and thoughts that we have any representation
or any knowledge of the world, and it is only by
collating, critiquing, and reconciling the vast
variety of our different accounts of the world
that we are able to form anything approaching
an adequate model or competent picture of it.
There is nothing dispensable about this fact.

MW: We are not saying that all objects of
    interest necessarily have an identifier.

Good.  It is very important to recognize the
practical impossibility, not to mention the
theoretical impossibility of doing this.

MW: We are instead saying something about how we will
    manage information about the objects, and that to
    be able to refer to objects consistently within a
    system, we will give them a system identifier (to
    add but no be confused with the arbitrary number
    of other identifiers the object is allowed to have
    through other parts of the model).  This is what I
    would call an engineering design decision, rather
    than an ontological commitment.

Maybe I can convey the importance of this if I explain that
all of the following brands of information involve the same
dynamic array of problems at this primary level of analysis:

1.  Sensory data.
2.  Measurements.
3.  Identifiers.

Yes, the precise values in each case are arbitrary in any absolute sense,
since they depend on sensor specs and aberrations, frame of reference,
history of the naming faculty or the hash code generator, scale of
measurement, topology of nearby space, and so on.  This is exactly
why "system identification" is such a non-trivial problem and
why we can only get a grip on ontology through comparative
and relative operations on the pile of bits that the data,
measurements, or identifiers would otherwise remain.

MW: We could have chosen to require implementers to use the general
    identification scheme for all identification purposes, but the
    consequence would have been that queries would take 10-100 times
    longer to run.

Sigh, if only we could get Nature to code better.

Jon Awbrey

o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o