Russia: Toward Engineering Semiotics
Decisions to create CIM/MIS enterprises should be based on the perspectives in IT and management science, considered not just in a short-term but in mid- and long-terms as well. The milestones that separate these three time scales are determined by particular events rather than physical time frames. In the short-term scale, basic technical solutions have been already found or actively explored, and the major task is to implement them. The respective basic notions have become clear or at least familiar. For the realization of primary short-term objective in development of new technologies, newcomers will have some advantages over old-style ERP implementations. An example is a “fresh” e-business technology in the greatest ferrous metallurgy enterprise POSCO (Korea). Realization of short-term objectives will require 5-10 years, by a rough estimate.
Achieving mid-range objectives will require the use of deeper models of system control compared with the old-fashioned MRP/ERP model. It would require more active involvement of designers in the analysis of development perspectives, and in accumulation of practical experience of model use (a rough estimate - 10-20 years).
Basic research required to reach long-term objectives is not completed yet, therefore it is difficult to give any estimates of time needed for their realization. We can only project the most important tendencies. Such "three-level" approach shows the logic shift in key problems of Information Systems development from concrete technical and technological aspects to more general methodological and ideological aspects. Whereas short-term and partially mid-term forecasts can be effectively used at an individual enterprise, the conclusions that follow from the other part of mid-range predictions, and especially from long-term forecasts assume actions at the state and international levels. But these actions imply the need to solve not just technical and technological problems, but also the social ones. The general analysis indicates, that the development of IT will require solutions of biological and social problems.
Despite of such a difference between time scales, there is a general direction in the IT development, which is the increasing unification and standardization of components. Short-term objectives can be achieved by direct technical and software-engineering methods, while achieving mid-term objectives would require standardization and unification of semantic elements as a basis for new technological level of B2B and B2C communications, and the long-term objectives should be based on viable systems concepts. Obviously, long-term objectives cannot be reached using traditional top-down regulation based on more or less arbitrarily selected standards that are weakly related to practice. Instead, “natural standards” have to evolve from a long history of trails and errors. Also, they should resemble "natural" information objects and processes that are universal for all viable systems . The strategy of IT development at the Magnitogorsk Mettallurgy Complex (MMK) (www.mmk.ru/eng/index.wbp ) is based on the criterion of “standard elements” , . Another important criterion that was taken into account was the speed of the IT development . Several generations of IT tools and facilities have been replaced in the MRP/ERP systems, but the basic set of functional elements remained standard . Therefore the facilities themselves are far from being stable. Based on these ideas, a practical conclusion has been reached that particular decisions in the IT area should come from considering perspectives of technology development, rather than from the point of view of the current conjuncture.
2. Short-term objectives.
The basic concepts and business-processes for the enterprises IT have been developed in the frameworks of the MRP/ERP model , and it is necessary just to adopt them more quickly. It becomes especially important because of increasing use of international standards in Russia. As a newcomer in the ERP world, Russia does not have obsolete technological traditions to overcome, thus it will be more easy to disseminate new technologies. Practical experience of implementation of Oracle HR module at the MMK has shown that there were no problems in accepting standard concepts and business-functions used in this package. These concepts are well defined and more precisely correspond to real features of HR management as compared with non-integrated technology. For example, Oracle HR provides a well-defined centralized support for multiple organizational structures and positions structures. Also it supports the “date-tracked objects” technology allowing to trace the history of positions, personnel attributes and assignments.
As for the “instrumental standard components”, the major vendors of IT actively implement OMG+W3C+Java standards (UML, XML, XMI, MOF, CWM, J2EE,EJBs,JMS ). For example, the Unisys actively develops the UREP technology (Unified REPository) with a unified data exchange between the MOF, UML and XML by the OMG’s XMI. These methods allow users to organize a new level of integration (e.g., through DTD for UML models, DTD for metamodel in MOF and DTD for Data warehouse metamodel CWM). At the next stage of UREP development, the Unisys plans to use the new versions of the OMG standards (UML-2 and MOF-2).
The Oracle expands a standardization of “elements” for business modelling too and together with Unisys and IBM participates in the XMI development. The Business Components For Java (BC4J) are used already in some new e-business suite modules (e.g., Self-Service, CRM). The BC4J are founded on the J2EE/EJB and XML principles. In the Oracle 9i the new OLAP technology that is based on the OMG’s CWM will replace the former Oracle Express . As the Unysys, Oracle works on a more deep semantic metamodel on the top of the JDeveloper which is based on UML, MOF, XML, and XMI.
As for OMG standards, there is an active work on a new UML- 2, which will be integrated with the base ISO 10303 EXPRESS language . The work on the MOF-2 continues as well. Additionally a new version of base XML document model DOM-2 is issued already. These developments give new possibilities not just for B2B, but for H2H (human to human) and A2A (application to application) knowledge exchange with common Web-based repositories and “ontologies”. But deeper semantic models are needed for better use of this potential.
Our MMK already uses major Web+Java oriented technologies from such vendors as Oracle and PTC. This allows us not just to unify the users interface but to use “superthin” clients when even the Web browser “turns” on a terminal server. Taking into account the above-referenced trends we plan to carry out a training on UML, XML, CWM and JDeveloper.
3. Mid-range objectives.
As it was said above there is a need for deeper models for enterprise data and activities similar to the ERP/MRP model. Although such inter-enterprise model extensions as Supply Chain, CRM, e-Procurement/Sales etc. are very hot technologies now, but they need deeper semantics. For example, the new EPISTLE Core Model data modelling methodology – ECM (http://www.btinternet.com/~chris.angus/epistle/ ) assumes that “Customer” and “Supplier” are not valid generic entity types because they are both “roles”, whereas generic entity types must represent the essence rather than roles. One of the important features of the ECM is using the 4D Ontology for object description that implements the 4th “time” dimension. The idea of using “date tracked objects” is very important for flexible products life cycle support (PLCS) and for the “biological” point of view to evolution of Products (“technetics” of Boris Kudrin) . I mentioned above already about some features of such date-tracked objects approach in the Oracle HR ERP module. But these examples are “exceptions to the rule” in ERP now . The 4D approach “demands” a global use of “date tracked objects”. Another example concerns inventory items modelling which is typical in MRP/ERP. Traditionally they are represented by catalogues with a unique surrogate identifier for each item. Further details in BOM are referenced via an inventory item (catalogue number). This may be good for complex assembling type items, but this approach is too rigid for non-assembling parts that may have multiple (or loosely restricted) combinations of their properties . As it was proved in biology such combinatorial flexibility exponentially “degenerates” into rigid hierarchy with increasing number of assembling parts. Therefore the problem is not important for complex assembly products but it is typical for “material oriented products” (rolled metals for example).
It may be sufficient for an enterprise to have “rigid” catalogues used in practice within that enterprise, but not for e-marketplaces where a more flexible approach is needed. Deep flexibility is needed for B2B and A2A communications both in the “horizontal” and “vertical” directions. A foundation for such flexibility was discussed long time ago in biology and was expressed explicitely in the “duality principle in a Classification Theory (CT)” in the 70th in Russia ( although the main publications are in Russian but there are some references in English - http://home.comcast.net/~sharov/biosem/time/time.html , http://home.comcast.net/~sharov/biosem/schreidr/schreidr.html , http://grouper.ieee.org/groups/suo/email/msg01920.html).
The main assertions of the CT are:
(I) Any Classification System has two Dual parts - "Taxonomy" and "Meronomy". The first one is "external" and connected with ordinary set theory relations (unions, intersections, hierarchy (a subclass of)) etc..
(II) The second one is "internal" and connected with Properties/Parts structure (archetype).
(III)CT differs from a hierarchy - combinational structure of Taxons and hierarchy - combinational structure of Properties. There are 4 extreme points of combinations of that two scales ( Hierarchy/Combinations of Taxons, Hierarchy/Combinations of Properties).
(IV)A strict hierarchy of Taxons can be described by pure combinations of Properties.
(V) "Good sets" ,their members and standard set theory relations are described by the "Taxonomy", but the dual part "Meronomy" doesn't fix the sets of objects in principle . Only the "subject areas" with "open" object types and explicitly defined properties for them. A "good" classification system must have the both parts but in practice very often only the taxonomy is used EXPLICITLY . And the Meronomy is “hiden" in human minds. The both parts are used in "Determinants” of biology (for example “birds nests Determinant") but for e-technologies there is a need in much more formalization and structuring of the Determinants.
(VI) The CT differs from a "subject area" or "classification field" . The first one is "not a closed" class. The last is a "good set" when the proper "primary" identifications from real objects to "minimal" taxons are already defined. The minimal taxons "substitute" real objects in any model. It is important to distinguish "taxonomical" properties from deeper "diagnostic" properties . A value of a taxonomical property may have a complex connection with them.
Classifications based on “pure” taxonomy are widely used for Product description for example for “upper” levels of description with rigid hierarchies of taxons. The Meronomy is not widespread now . Our experience in using the dual approach for describing products at the MMK showed a great potential flexibility of such description. We used Determinants for “manufacturing classes” of products , “cost classes” and “pricing classes” which were formalized for automatic identification of “individual” objects as members of proper classes. This experience was only a first step in this direction. There is a need in deeper analysis of different types of values of Properties. Such analysis is related with some theses of the Measurement Theory (MT) - http://www.upa.pdx.edu/IOA/newsom/pa551/lecture1.htm and combines it with CT ideas.
This direction has deep semiotic roots and will play an important role in the mid-range and long-term IT developments. But may be it will be better for this to replace the “meron” concept in the “Meronomy” part to the “property” concept . Indeed the “fathers” of the duality principle understood the “merons” wide as "morphological" and "ecological" features. Already in the ODMG-93 standard there was defined two types of properties "attribute" type and "relation" type (ecological features). Following this direction it will be natural to regard the parts as “properties” too (morphological features) taking into account the (MT) uniform platform for quantitative and “nonquantitative” measurements. Suppose this more deep foundation is needed for the core notion of the Semantic Web – “triples” as well.
At MMK we already used the EPISTLE Core Model, “duality principle” and some above mentioned ideas in our runtime meta repository. Although it was only a prototype work but it allowed us to support not just the first order logic but the second order logic as well.
Activities in developing deep enterprise models are related with a wider use of the system approach (for example ISO 184/SC5/WG1 - http://www.mel.nist.gov/sc5wg1/) and the Stafford Beer’s 5-level model of an enterprise as a living organism . The TOC model ( www.goldratt.com) is very close to the 3rd and 4th levels of the Beer’s model and it was already announced that the MRP III (Money Resource Planning) will use the TOC model.
New models require deeper semantics and there are active developments in this area now. Except the above mentioned activities on the OMG, the W3C’s “Semantic Web” direction with it’s RDF+DAML+OIL + DAML-L foundation is a mainstream as well. Other projects include the ISO 184 SC4/WG10 Data Architecture (AP221, IIDEAS – http://www.tc184-sc4.org/ISO18876/index.cfm) and the IEEE SUO (http://suo.ieee.org/) activities. This situation with multiple “points of growth” is typical for early stages of development but further progress would require development of some “standards” mentioned at the beginning of the paper. There are already some efforts to integrate different models - http://citeseer.ist.psu.edu/context/1823876/706748 etc. but to be successful this integration should take into account deeper semantic foundations than are used in current activities around the ontologies. Some of them were referenced above and in the “IT Strategy for the 21st century” paper  (a complete machine translation into English can be found at the end of the message - http://grouper.ieee.org/groups/suo/email/msg01920.html) . The core of modern Semantic Web activity is around the Product and Service Ontologies which are at early stages of development (http://www.heppnetz.de/). There must be done a great deal of work to transit from B2B to A2A with minimal taking part of slow and error-prone “human parts” .
It may be not the best strategy of some IT-communities to formalize models “up to the end”. Apparently, comprehension of the necessity to “the human factor” explicitely in the models will require long time and serious efforts. There is a need to turn minds back to the Peirce's semiotics with it’s “syntactique” , “sémantique” and “pragmatique” and to modern activities around it. Only one example of such “semiotics” description of “pressure” property of such physical object as pump. From the syntactic point of view a “Format” is determined for the pressure. From the semantic point of view a “quantitative” measure scale is determined for it with proper units of measure. From the pragmatic point of view some "ordinal" scales of estimation may be defined for the pressure with such values for example as “low”, “normal” , “high” . At last the relations between the semantic scale values and pragmatic scales values must be defined. The pressure property can be generalized for more wide classes of objects in the Repository. From the semantic and pragmatic points of view such generalizations may be different . Such from the semantic point of view the generalization can be run up to “pressure in liquid and/or gas” but for the pragmatic point of view there is a need in much more details.
From more wide Pragmatism point of view the ontology itself refers to what humans have agreed to call "existing" and what can be operated following the same rules as formulated in our language. We separate mind from matter for communication purposes ( http://home.comcast.net/~sharov/biosem/txt/umwelt.html). New models must take into account this “relativity”. A possible approach could be to include “analytical agents” as black boxes into the models and then to distinguish them from the “constructive agents” . It is only a “mark” of the direction but it has close connection with the VSM of Stafford Beer - http://www.ototsky.mgn.ru/it/beer_vsm.html.
He used such “black boxes approach” for the “algedonic” universal control chain. Another very important feature of the VSM is its “integration” of Systems 1-5 where System 2-3 integration correspond to the MES-ERP integration (for the enterprise level of the VSM “recursion” ). There are another very important features of the VSM which are not used now in CIM/MIS products and which must be used in mid-term perspective .
Although the Web provides vast possibilities for consensus in the “semantic plug-and-play" components , but the ambitions and habits of various research groups and the existing infrastructure of social institutes may postpone their realizations.
4. Long-term objectives
It is not easy to make predictions over more that 20 years, nevertheless some forecast are possible. I expect an increasing “humanization” of the models on the basis of integration of the Stafford Beer’s “Viable System Model” with the Maturana and Varela’s “Autopoietic Systems”. The role of biosemiotic ideas in IT would increase. Indeed such “objects” as musical scores, blueprints, computer programs or DNA molecules from the biosemiotics point of view have an information "micro-level" nature and the biosemiotics distinguishes it from "macro-level" of "structure characters" . It differs an "algorithm of activity" from "activity" and "results of activity". There are principle different types of activities to create the "micro-level" objects, to copy them, to use them for "constructing" a "macro-level" object etc. ( http://home.comcast.net/~sharov/biosem/txt/biosem.html ) . Other very important concepts of “semantic closure” and “metasystem transition” ( http://home.comcast.net/~sharov/biosem/geninfo.html ) will be taken into account as well. I expect an integration of the Autopoietic systems model with biosemiotics and biohermeneutics (http://home.comcast.net/~sharov/biosem/chebanov/enlog.html).
The practical use of “engineering semiotics” will increase. Currently it is at the beginning stages of development. As it was stressed in , the need of social changes (following the Norbert Wiener’s “Cybernetics and society”) will be more important. And it is connected not just with need to have a consensus in IT development but with the survival of the mankind. There are many developments in this direction too. For example the “Principia Cybernetica Project”, the “Global Brain” project, the “Digital Government” project etc. But discussion of these topics would require a special paper.
Stafford Beer. “The Viable System Model: Its provenance, development, methodology and pathology” : http://www.echelon4.com/content%20files/VSM.pdf
L.Ototsky, V.Savin. “Seven criteria for choosing an ERP system in Russia” “Open Systems Journal”, 1998,#4-5 (Rus) http://www.osp.ru/os/1998/04/179564/
L.Ototsky. “ A strategy for choosing an ERP system for MMK”, “Oracle Magazine (Russian Edition)” , 2000,#6 (Rus)
L.Ototsky. V.Savin. “A thorny road to modern technology of management” “Open Sytems Journal” , 1998,#2 (Rus) http://www.osp.ru/os/1998/02/179412/