metapattern meets RM-ODP
Pieter Wisse
Metapattern is a powerful approach to conceptual information modeling. It is
not a comprehensive development approach for information systems. Therefore, it
must be combined with other metamodels, methods, and so on.
The idea behind combining forces, or forming an alliance, should always be
this: that a particular task must be fulfilled but is too big for a single unit
to complete successfully. When several units contribute their efforts, a task's
critical mass may be, literally or figuratively, overpowered.
Success by alliance comes at the price of coordination, though. This implies
that an alliance is never totally homogeneous, as units contributing similar
efforts would remain, by definition, uncoordinated. Many people can attempt to
lift a rock. If their aggregate energy more than equals the rock's weight, then
it is only when they apply energy simultaneously that the rock will move.
Making people lift on the count of three, for example, would be a coordinating
effort.
At the other extreme of typical alliances, as shown in Figure 1, all efforts by
units are completely heterogeneous. Coordination becomes prominent. Considering
the characteristics of participating units, a task is decomposed into disjunct
aspects. For the completion of the overall task, actual efforts are matched to
such aspects.
Figure 1: Spectrum of alliances.
In practical life, alliances lie somewhere in-between. It depends on the requirement of coordination as the only different effort and, on the other hand, of all efforts being different. Looking at this range, it should be evident that alliance and organization are synonyms. The term alliance helps to stress the importance of contributions, while the term organization has become associated with the (static) structure of participating units.
I have purposely called participants of an alliance units. A unit
suggests autonomy, making clear that every alliance exists because of
a fundamental tension. The assumption is that, given the opportunity, a unit
would rather execute the task by itself, exclusively reaping the rewards. An
important aspect of coordination, therefore, is to counteract hegemony.
Ideally, units exhibit self-control. Aware that their particular efforts are
insufficient for task completion, they recognize the contributions of other
units. Each unit wants its autonomy respected, relinquishing in turn some of
its original autonomy. It accepts coordination for the benefit of enjoying a
fair share of success which would escape it without contributing to the
alliance.
The second reason to use the term unit is its generality. It covers a wide
variety of phenomena. A unit may be an individual person or an organizational
unit. Unit can also be given a less traditional meaning, such as that of method
or paradigm.
The latest developments in the philosophy of science acknowledge that the
complexity of reality cannot be explained or changed according to a single
paradigm. This has led to acceptance of multiparadigmatic approaches.
No paradigm (unit) may claim absolute priority. The proper alliance of such
units supports the successful achievement of the specific, often complex
(scientific) task.
An application of the metaparadigm of multiparadigmatically coordinated efforts should be the development of complex information systems. Such development is never mere deployment of scientific results achieved elsewhere. It follows from the need for innovation that a development process has scientific aspects, too. Without a scientific attitude—healthy curiosity—it would be impossible to gain sufficient understanding of the problem, opportunity, or situation. It would be equally impossible to design, construct and implement an optimal information system as the solution.
A scientific attitude alone, though, could be disastrous. For example: if
the task is to improve the quality of service in a government agency. Other
aspects need to be recognized and treated accordingly.
Suppose that a particular unit would perfectly match some aspect x of the task,
but would be unsuitable for aspect y. And suppose, too, that the unit
one-sidedly declared its paradigm valid for both aspects x and y. (This often
happens with the best of intentions: therein even lies the danger.)
Such a course of events is known in philosophical discourse as reductionism, or
simply reduction. What happens is that understanding, and subsequent practice,
of y is reduced to the theory and practice underlying x. The difference between
an open alliance and reduction is sketched in Figure 2.
Figure 2: Reduction of conceptual modeling by construction orientation.
Of course, reduction can work in any direction. It is equally harmful to view tool construction, including characteristic precision, as if the situation for its intended application must be charted. Such opposite reduction, when compared to Figure 2, is shown in Figure 3.
Figure 3: Reduction of construction by orientation at conceptual modeling.
Reductionist practices in this sense are normal in human society, and not
always bad. Someone visiting a doctor expects to be treated as a patient. But a
doctor who does not meet the patient as a person pursues professional reduction
much too far—that is, she or he is not really a professional.
The saying that "To a man with a hammer, everything looks like a
nail" expresses the practice of reduction well.
A major issue of task coordination can now be restated. An alliance can only
remain successful when reduction by any participating unit is avoided or, when
that is discovered as too ambitious, kept at an acceptable level. Ideally,
every unit combats reductionist urges itself.
Coaches of top performing sports teams consistently select players based on
their strengths. One such strength is the ability to avoid exhibiting eventual
weaknesses. Thus, a team may be considered a combination of participating
strengths.
Development of a complex information system is like a team effort. Units should
be included on the basis of how well they match particular aspects. Any
weakness concerning other aspects do not matter, as different units will
properly match them.
A variation of this theme is that one unit must be completely knowledgeable about all other participating units. It is important for their task-oriented efforts to be optimally coordinated. This makes it both necessary and sufficient for a particular unit to know how to interact with other units. It can, and for the sake of effectiveness and efficiency should, be totally ignorant about how another unit internally acts to produce such contributing efforts.
To develop a fitting information system, it is necessary to learn about the
purpose(s) it has to serve. This entails acquiring an understanding of reality.
Such understanding should be as rational as possible, in the sense that reality
is viewed as essentially structured.
What having a structure (or its equivalent, being a system)
assumes is that elements exist, with relationships between them. Special status
is awarded to the element of the environment or, to be more precise, the rest
of reality.
A strength of metapattern lies in its support for more realistically
understanding reality. It escapes the traditional limitation of direct
synchronization between the concepts of element (called real object, or simply
object) and its behavior. Usually, a one-to-one correspondence is defined. As
this correspondence is implicitly assumed, it is difficult to escape.
With metapattern, an object may be attributed with a multitude of behaviors
(intexts), with each behavior unambiguously defined for a particular context.
Metapattern includes a simple method for integrated visualization of multiple
contexts, of partial objects each having its own behavior or intext, and of
integration of partial objects to constitute an overall object.
The principle of contextual differentiation of object behavior opens the
possibility to reflect comprehensively on greater complexity of reality. Such
conceptual models may be easily documented and communicated using metapattern's
simple notation.
The emphasis on conceptual modeling as a strength is a clear indication of
metapattern's weakness. It is not a paradigm, or metamodel, to support
predominately technological aspects of (digital) information system
development. In full recognition of this weakness, I advise not
applying metapattern beyond the area of its evident strengths; that is, outside
conceptual modeling. Otherwise, aspects requiring a characteristic paradigm for
their optimal development would be reduced to what metapattern allows. With the
focus of metapattern on (life in) reality, the equally necessary focus on tool
technology would suffer.
With metapattern optimally matched to conceptual aspects of development, one
or more missing units are needed to successfully complete the overall
development task. What are suitable candidates? Does a unit even exist that
shows strengths for all aspects, including conceptual modeling? Such a unit
would make it unnecessary to let metapattern participate in development.
The number of methods, paradigms or metamodels (call it what you will) offered
as employable units for information system development has grown beyond
reviewing capacity. Here, instead, only one unit is considered for alliance.
This serious candidate must be introduced first.
A pervasive characteristic of western society is its variety, occurring for
reasons that range from free competition to state intervention. Competition
creates (some) variety, offering consumers at least the illusion of choice.
Historically, when a state intervenes, it is often without regard for resources
elsewhere. A famous case involves different widths between rails, causing
carriages to be transferred at some international borders from one chassis to
another.
At any rate, variety is a way of life; it’s a reality. Information processing
is no exception. As with a train strictly keeping to a proprietary track,
information processing remains relatively simple when performed in isolation;
that is, using stand-alone resources. Increasingly, however, resources are
combined for information processing, as a counter-phenomenon to variety exists,
exhibiting itself in cooperation and communication. Adding to the complexity is
that many such resources have not been developed with an alliance of different
units in mind. (The recurring theme of alliance is, we hope, not lost on the
reader.)
Accepting the fundamental variety—even honoring it where it supports special
requirements and/or promotes innovation—the question becomes how distributed,
often heterogeneous resources can still be successfully combined for
information processing.
Throughout the ages, the answer has been: establish an agreement on exchange
format (Figure 4). Such agreements are commonly known as standards. Again,
digital information processing is no exception.
Figure 4: The principle of standardization.
Standards originate from several sources. Official, or de jure,
standards are issued by Standards Developing Organizations (SDOs) in which
nations participate, usually represented by their respective national
standardization bodies. Actively involved in international standards
for information processing are the International Telecommunications Union
(ITU), the International Electro-technical Commission (IEC), and the
International Organization for Standardization (ISO). ITU and ISO have a
long-established healthy tradition of adopting each other's standards. More
recently, ISO and IEC have joined forces as far as information processing standards
are concerned; they work together through Joint Technical Committees (JTCs).
Over the years, many standards have been developed to combine resources over
distance (telecommunications), and/or to bridge technical heterogeneity. A
standardized framework for information processing—a metastandard—has been
developed by ISO/IEC. It is called the Reference Model for Open Distributed
Processing (RM-ODP). It goes without saying that RM-ODP is oriented at distributed
information processing. Otherwise, there would be no need for a standard. The
standard's objective is for information processing to be open; that
is, no obstacles for successful information processing will remain when all
participating units conform with RM-ODP. It is, so to speak, the standard which
secures the opening required for successful exchange between resources
that may otherwise be different from each other. (An alternative name for
RM-ODP might be reference model for allied information processing resources,
which directly illuminates the critical issue addressed by the standard.)
The fact that official standards have a predominantly technical orientation
is an early indicator that metapattern and RM-ODP could complement each other.
A related example would be a simple telephone connection; an agreement must be
made about how to transfer elementary signals while preserving (more
accurately, reconstructing) their original structure to represent higher-level
signals. Not standardized is what such higher-level signals, or information,
mean. The vocabulary of the conversants is also not subject to standardization.
Again, it is the signals, at an intentionally meaningless level from the (end)
user perspective, which are standardized.
It is this limitation—avoiding any attempt at standardizing interpretation by
(end) users—that has spurred technological development. It is the perfect
example of a successful alliance. The aspects of tool use and tool development
meet when base signals are transferred.
A note of warning is in order (see different meanings of object, below, for
more). Aspects of communications such as phonetics, syntax, semantics and
pragmatics are relative. From a particular perspective, they all have a
corresponding meaning. Thus, pragmatics are also relevant when applying a
strictly technological focus. Pragmatics with a life focus, however, mean
something different. At the minimum, levels are shifting. It seems reasonable
to suggest that, for example, a tool's internal semantics are in many ways its
user's external phonetics. Semantics, therefore, within a tool such as an
information system, is a different concept from semantics as experienced and
practiced outside by the tool user. Even though it is a simplification,
thinking in terms of restraining descriptions of communication aspects by
focus—and only then investigating how they might correspond—may eliminate much
confusion between collaborating units. Figure 5 shows a general example.
Figure 5: Possible shift of communicative (semiotic) categories between different focuses.
Figure 5 raises awareness that some communication or behavioral aspects are impossible
to translate between focuses. This cannot be changed; it’s just the way it is.
Such fundamental differences should be fully exploited, not denied—and that is
why essentially different methods, metamodels, or whatever they are called, are
required when a complex task may be characterized by a multitude of focuses.
Success calls for an alliance of metamodels.
Again, meaning should not be reduced to a single focus. Such familiar use of
similar terminology has unconsciously obstructed communication, rather than
promoting it, as meanings can differ radically. Insufficient awareness of real
differences often makes units believe they are competing while, in fact, they
could act as allies to their mutual advantage.
Earlier, I labeled the concerns of standardization bodies such as the ISO as
primarily technical in nature. Does RM-ODP conform to this idea of a technical
standard, even though it is a metastandard?
Two answers are possible. The first takes a point of view outside
RM-ODP. For example, if viewed from the conceptual perspective of metapattern,
the technical nature of RM-ODP is obvious; the standard primarily addresses
concerns of construction, implementation and operational management of
distributed tools for information processing—concerns not addressed by
metapattern. An alliance of both metamodels, therefore, can make complex
development tasks feasible.
It may look different, however, from inside RM-ODP. It is not so much
the claim that the information viewpoint is completely covered by the standard.
Being a metastandard, an alliance could well consist of metapattern supporting
development of information views. In fact, this would be a real opportunity,
except for RM-ODP insistence on the use of basic modeling concepts that show
weaknesses for conceptual modeling when compared to metapattern.
A short overview of some of RM-ODP's basic modeling concepts may suffice to
appreciate the difference with metapattern with regard to conceptual modeling.
What is reality according to RM-ODP? First of all, reality is considered a
universe of discourse. And "[t]he elements of the universe of discourse
are entities and propositions."
An entity is defined by RM-ODP as "any concrete or abstract thing of
interest." This would make an entity, or its synonyms, thing and element,
the same concept as metapattern's (real) object.
A proposition is "an observable fact or state of affairs involving one or
more entities." This (last) sentence itself is a proposition, too. The
ontology of RM-ODP, however, is difficult to extract. It seems that the
"universe" is made up of entities, while the "discourse" about
it occurs through propositions. Metapattern tries to separate views of reality,
on the one hand, and views of reality's representation (information) on the
other.
RM-ODP defines an object as "a model of an entity." With object
equivalent to entity, in metapattern terms the question must read: What is the
model of an object? The answer needs an introduction.
When a real object (entity in RM-ODP) is observed from a behavioral
perspective, variety must be acknowledged. Metapattern states that an object is
presumed to be an overall object, consisting of partial objects, each
attributing with different behavior. Now, what is the model of an overall
object? It is a configuration of information objects, specifying multiple
contexts as required to accommodate diverse behaviors; the partial objects are
represented by unique nodes in this configuration, connecting every context
with a corresponding intext which specifies the partial object's static and
dynamic behavior.
But doesn’t RM-ODP accomplish the same result through naming contexts? It
defines a name as "a term which, in a given naming context, refers to an
entity." What multiple naming contexts do solve are problems of synonyms
and, in particular, homonyms. They do not directly support, however, finely
grained differentiation of behavior.
What about object interfaces in RM-ODP? Are they not equivalent to
metapattern's contexts? To some extent they are. It could be possible to
express an interface in terms of other objects (metapattern: information
objects) and their relationships. A disadvantage of this interpretation is that
the concept of interface is no longer available to support different
implementations of behavior within a single metapattern-based context. It would
thus be wise to keep the technologically oriented meaning of RM-ODP's concept
of interface intact.
Whichever way interface is interpreted, RM-ODP's weakness in conceptual
modeling, when compared to metapattern, should be apparent from a short
comparison. A full review would be far beyond the competence of the author, anyway.
In general, a professional conceptual modeler should retain a balance between
knowledge and ignorance of technological issues.
From this short discussion it may be observed that the use of the term
object is typical for the shift that, by definition, occurs between focuses.
Applying RM-ODP's terminology, object seems a pragmatic concept, corresponding,
as it should, with RM-ODP's characteristic focus on tool technology. Metapattern
does not even recognize that technically pragmatic concept as it concentrates
on conceptual modeling. It puts forward the concept of information object to
enable translation of conceptual models onto construction models. As such,
metapattern's information object may well be labeled an intermediary concept,
too. It paves the way for the specifications of tools which, when constructed
and made available, are integrated into the whole of reality.
Within metapattern, object is also a pragmatic concept. Indeed, terminology is
highly isomorphic. But according to metapattern, what is meant by this is an
object behaving in the totality of reality. This is different from RM-ODP's
object, which behaves within a particular information system.
Ideally, pragmatics of real objects are exhaustively mapped onto pragmatics of
digital technology objects. But such a full representation is illusive. The
Belgian painter Magritte (1898-1967) humorously exposed the core of the
insoluble problem when he included the statement "this is not a pipe"
in a painting in which he figured a pipe. Thus, in a metapattern-based
conceptual model, the term by which an(other) partial object is named should not
share the nil identity with that partial object. Instead, the reference to the
term must be considered a property of the partial object named. Otherwise,
confusion sets in between the proper object and its representation. It is not
the name proper which is an object's property in the sense of complete
ownership; it is one object having or given the use of another object for
its name. Remember that what is modeled concerning an object's name is,
again, a representation. Out there in reality, the name is also an object.
It is difficult to illustrate the basic incongruence of real object and
representation, but the reason for this problem is quite simple: any figure or
model, would, by definition, imply the appearance of the real object through a
representation. An opposite problem caused by the nature of a model: infinite
regression rules the representation of a representation. Therefore, a model of
how a representational object stands to the real object it represents needs to
be simplified. In Figure 6, a choice is made for the real object 1 to be
interpreted as a representational object. (The “real” cube represented must be
understood to be missing by default.)
Figure 6: Incongruence between object and representation.
With real object 2 considered a name, it may be applied to represent real object 1. It is precisely this act of assigning representation which creates both relationship and distance between the two real objects involved. What their relationship amounts to has been extensively theorized about by disciplines such as semiotics. Now, with metapattern, their preferred distance can also be formally expressed. As indicated above, it follows that a representation of a representation should be excluded from sharing the nil identity with the representation of the real object represented. To avoid behavioral confusion, they must not participate as partial objects in the same overall object. This is shown in Figure 7.
Figure 7: Constraint on the relationship between real object and representational object.
Without such understanding, it is difficult to avoid substituting, rather
than simulating, the behavior of the representational object for that of the
real object. Starting the analysis from higher-order representations increases
the reduction of reality even more. Is substitution, and its danger, not
exemplary for technocratic approaches to reality?
Conceptual modeling is least reductionist when based on, and performed using,
an ontological, reality-oriented system of (meta)concepts. A
technologically-based metamodel can never be made to express what it different from
its origin. RM-ODP, for example, can hardly be considered a metamodel of
reality. All it presumes to exist are entities (things, elements). Structural
elaborations are grounded in the tool technology. That focus must be
complemented.
Only when the name-as-real-object is distinguished from the name's
representation in the conceptual model is it possible to appreciate that, in
reality, the object-object and the name-object are ontologically different
objects (to use just two examples). They are representationally related, which
is important. But, to repeat, it does not qualify them to be considered partial
objects of one and the same overall object. Magritte made this perfectly clear
long ago in his brilliantly immediate way. For C.S. Peirce (1839-1914) it was
the principle of his semiotic philosophy.
Denying the fundamental dissociation of an object and its representation is
also at the root of branches of, say, idealized artificial intelligence, which
aims to equal and even surpass human intelligence. The answer: there is nothing
artificial about intelligence. Instead, every real object should be considered
to have its own, characteristic intelligence. Therefore, digital machine
intelligence is a much better term. It acknowledges that a machine may be intelligent
but essentially different from a human being.
Important choices about necessary limitations should be made through conceptual
modeling. When conceptual information models are forced within a basically
technical frame of reference (metamodel), it is almost a certainty that choices
will never be optimal. This reduction is not yet fully evident with information
systems for control of closed, technical systems. The unambiguity of technical
system behavior is even, so to speak, preordained. Thus, there are no
fundamental differences between the behavioral domains of technical system and
the controlling tool. The behavior to be modeled is the technology. (And that
is why traditional object orientation is successful for information systems
controlling technical systems.)
But a tool for use in an open, social or socio-technical system cannot escape
the reality of different behavioral domains. Representational objects in the
behavioral domain within the tool itself can simulate real objects in
the behavioral domain of the wider social system, but they can never substitute
for them. With substitution a pure impossibility, the range of possibilities
for simulation indicates the risk associated with confusing real objects with
tool-based objects.
One of the conclusions from this philosophical analysis should be that RM-ODP's
basic modeling concepts may still operate satisfactorily for conceptual
modeling of closed, technical systems. Then, the simple, one-to-one
representational correspondence between real object (RM-ODP: entity) and
information object (RM-ODP: object) will often not constrain what is minimally
included of a conceptual model.
The expressive power of RM-ODP's basic modeling concepts will often be
insufficient, however, for conceptual modeling of social systems, characterized
as these are by qualitatively greater variety and complexity of (real) object
behavior. Metapattern's richer language invites the modeler to respect
behavioral diversity of a real object, resulting in conceptual models with a
closer simulation fit. In a good alliance, RM-ODP's basic modeling concepts
should not be made to also conceptually model reality. They can be put to
effective use starting with the creation of a tool-oriented construction model
from the conceptual model created earlier with metapattern.
The obstacle which RM-ODP has formally (with conformance requirements part
of the standard) erected against a successful alliance can, in principle, be
easily removed. At the same time, it is difficult to achieve, as it needs a
change of perspective from inside to outside RM-ODP. The requirement for some
basic modeling concepts to be universally applied in all of RM-ODP's viewpoints
does not serve any real purpose. Actually, it risks reducing the
quality of overall development by forcing the extension of technical
interpretations on conceptual issues. This is a mistaken approach, encountered
in attempts to extend the metamodel of software engineering to conceptual
modeling. The isomorphism between what is seen with a focus on life and on tool
technology, respectively, exists only with a single focus. A dual focus (or
multi-paradigmatic approach) reveals a superficial isomorphism at most, to say
nothing about semantics or pragmatics. A small conceptual isomorphism is always
insufficient to deal with issues which are qualitatively different by
nature.
Again, reduction, in its philosophical sense, is difficult to defeat. It is
seductive, and it seems easy to transfer success with one aspect to another.
Why trust another unit to contribute? Why spend the extra effort on
coordination? And why share benefits of success? It often takes an experience
of failure to realize the benefits of alliance.
A standard is achieved rationally, and with the maximum guarantee of quality,
when concepts can be successfully mapped between aspects (or focuses), one to
another, or one to a configuration of others.
When the strengths and weaknesses of RM-ODP are explicitly recognized,
metapattern may inspire some improvements in the next version of the ISO/IEC
metastandards.
The idea of multiple contexts lends itself to distributed information
processing. What must be established is, if not an unambiguous procedure, at
least a heuristic for mapping conceptual diversity—and the opportunities
provided by it—to the relevant (distributed) configuration for information
processing.
Other relevant issues are metapattern's “standardized” conceptual mechanisms
for time and validity control. With respect to time, for example, RM-ODP
applied the epoch concept. It should be no problem to refine it to the level of
control suggested by metapattern.
What also must be developed in detail when an alliance is pursued are mappings
between metapattern and the schemata required by RM-ODP's information
viewpoint. In many ways, what are called invariant, static and dynamic schemata
in RM-ODP are combined by metapattern. One question is: What is still missing,
if anything?
In preparing for and eventually evaluating and alliance, the overriding
awareness should be that success comes through building on genuine strengths of
participating units rather than through compensating weaknesses.
Metapattern must be open to continuous improvement. It is hoped that
professional conceptual modelers will be inspired to make valuable
contributions.
This text was included in Metapattern: context and time in information models (Addison-Wesley, 2001, see Appendix B).
2000, web edition 2009 © Pieter Wisse