User Tools

Site Tools


DMN Development and Progress

This page covers the activities and actions of the DMN RFP work…

RFP F2F3 meeting, London, 6Jan2011

  1. New material (eg Ron Ross's Q-Chart paper) has help clarify some of the statements we need to clarify in the RFP:
    1. Decision Logic is the goal of DMN; Decision Process is related but not part of DMN
      1. Decision Process can be that which requires a decision to be made eg a Business Process using Decisions, or a Decision Modelling Process that is the process by which the decision logic is defined / refined / prepared for automation
    2. SBVR compares with DMN: DMN is about instantiating a fact model (making decisions) based on conditions
  2. Examples of decision notations also include: KPI Decision Model and BRS Q-Chart (both TM by their authors)
    1. Enough similarities of the information used to make a shared metamodel and notation specification possible
  3. “Decision” is -not- a behavior per se, but a logical construct - behaviors can be associated with facts
    1. Decisions define facts, facts drive processes, actions and behaviors generally
  4. Metamodel for DMN will allow for consistency checking / verification etc, allowing metadata for exhaustiveness, exclusivity, (development) completeness, intention (ie traceability to policy / reference policy and business rules), supporting the notation(s)
    1. Traceability is important but generic / customisable - eg see also BRTI Business Rules Traceability Initiative from 2001
    2. Related to “single sourcing” (see BRS Q-chart paper) ie traced to single source of business rules, policies etc
    3. Metamodel operations are typically carried out by tools: eg redundancy checks
  5. Business versus Executable Decision Models: important to consider both (as in BPMN) although this is a challenge (as in BPMN: annotating business flows to be executable workflows is a non-automatic process)
    1. This is also a need for the “levels” of models (ie of differing details)
    2. cf “layers” which are of detail / abstractness eg subdecisions
  6. Classification - needed to identify notations covered (and not) potentially and in reality by DMN
  7. Fact Model - needs to hooks in metamodel eg for SBVR (business-level tools), UML Class (UML tools) etc
    1. For Process Decisions where a process data is needed, the vocabulary can include the BPMN model ie which task is complete for what entity etc
  8. Use Case types:
    1. Business Decision model + map to rule / decision engine
    2. Business Decision model + map to process and rule / decision engine
    3. Process Decision model
    4. Case Management Decision model
  9. Use Case domains:
    1. manufacturing
    2. healthcare
    3. telco
    4. CRM / marketing

OMG TM, Santa Clara, 6Dec2010 and discussion 7Dec2010

  1. RFP did not persuade AB what DMN was and its need
  2. Need to emphasise / add:
    1. business level standard
    2. DMN cannot be handled by SBVR, BPMN etc
    3. clarify NOT a new language ie references / uses existing or subsets of existing languages
  3. Scope needs to be better defined / narrower
    1. explain DTable role vs other metaphors
    2. what is needed re notation, metamodel etc
  4. Timing:
    1. LOI after DMN day
  5. FAQ appendix
  6. 1st page problem statement

OMG TM, Cambridge MA, 20Sept2010 and discussion 21Sept2010

  1. KPI DM presentation
    1. also links to BMM for metrics
    2. rule classification not used (although had been earlier) eg behavioral vs definitional
    3. patent issued for their particular way of modelling
  2. DMN RFP discussion: RFP should have
    1. DMN classification of decisions should be presented in the RFP
      1. this would be the structure of the decision dependency network / goal network, and the relationship to decision model types such as decision tables
    2. definitions and glossary
    3. difference between business and IT DM via the vocabulary
      1. need to translate from eg SBVR to eg UML Class / Java etc
      2. CIM vs PIM: decision is CIM ie technology neutral
      3. layered model: move from CIM to PIM as more detail added
      4. declarative nature matches SBVR declarative nature
    4. non-normative transformation examples eg DMN to RFP
      1. via example use cases etc
    5. inference network vs dependency network
      1. decision model could be defined “static” ie as a model, or could have its links defined by inference
      2. multiple model types should ideally be covered
  3. existing DM / contributions to frame DMN
    1. Donald C had experience of the dependency nw and its success
    2. KPI DM should also be defined as a use case
    3. RFP should have an appendix of such use cases explaining the pertinent points of the models used
    4. above should use a standardised format (TBD)
    5. use case contributions should be from industry experts, vendors, consultants in this field
  4. issues for RFP editing to be covered at future RFP edit sessions
    1. relationship with BMM
    2. glossary for DMN RFP to include the dependency network
    3. option or requirement for graphical notation for dependency network?
      1. eg KPI model is a profile of a std dependency nw?
    4. decision classification - example classification - to explain the wider scope of decisioning and the subset required by the RFP
    5. use cases appendix and structure for describing them
  5. F2F3 should concentrate on decision network and BMM relationship requmts

RFP F2F2 meeting, London, 5July2010

  1. Update on OMG MN meeting
    1. Vocabulary and terminology
      1. SBVR relationship: SBVR primarily provides a vocabulary (standard-based) although it also includes business constraints which can be used in decision models.
        1. SBVR goal for decisions models would be traceability from decision to business
        2. SBVR rules could be considered as “guidance” for decision development and/or business documentation to assist in decision development
      2. DMN goal is to allow use of “industry vocabularies” in whatever form is relevant - eg text, vertical-standards (SBVR, OWL/ODM, UML Class, XSD) or horizontal-standards (XSDs for domain standards eg MISMO, ACORD) (note for RFP)
        1. Consideration should be made of moving between vocabularies eg from SBVR to UML Class for a decision model to “make it executable”, or from an XSD-based model to an SBVR model “as domain standards convert to SBVR)
    2. Process models
      1. BPMN mappings can take several forms
        1. the Business Rules Activity (BPMN2)
        2. the process gateway
        3. some BPMN subprocess representing a decision tree or graph
      2. Use Cases are:
        1. “Rule/decision discovery” - Extract DM from a BPMN model for validation, find missing rules, re-use in other parts of the model
        2. “Rule/decision reuse” and reference - Embed a DM inside a BPMN model
        3. Same but also convert to BPMN notation
        4. DM “ruleflow” or embeds some decision process
      3. Not known - what is the precise delta between a BPMN-defined decision model (eg for tree or graph) and a decision model version…
      4. Motivation: bring BPM and BR communities closer with better understanding and skills *(note for RFP)*
    3. Formats: will depend on scope…
      1. Metamodel - per BPMN, DMN will have a metamodel of type CMOF (Complete MOF) or EMOF (Essential MOF) - see - probably submitters should choose with a justification
      2. DMN tools could import/export the following *(RFP note)* with some of these possibly being optional in scope
        1. drawing notation (metadata) eg colors used etc
        2. transformation (metadata) eg constructs allowing use of multiple metaphors
        3. content eg decision table columns, rows, cells
        4. vocabulary (metadata) eg source for definitions of content, types, etc
  2. Wells Fargo use case discussion
    1. *(RFP note)* introduction v useful for positioning DMN
    2. Use case: find rules in processes through observing repeated patterns and identifying case for externalising (eg frequency of occurences, frequency of change to these occurences, …)
    3. Use case: DMN as an “executable rule format” - this positioning is distinct from the “business modelling” focus of DMN to date, with the assumption that vendor tools will provide the DMN/metaphor - to - executable mechanism, but might just be “overall effect” vs “DMN” viewpoints…
    4. Use case: there are several levels in mapping from a BPMN process model…
      1. the gateway or activity
      2. a decision process: activities invoking decision models and any gateway that uses them
      3. a decision model or rule
    5. Examples provided:
      1. *(RFP note)* Decision model will need to cover visualisation options eg table ordering, etc
      2. Fig10 etc is an example of a Karnaugh map - - which is a “spreadsheet-type” model with conditions in rows and columns
      3. Fig12 had implied rather than defined ranges - this is more likely a model deficiency / reporting simplification rather than something that would be encouraged?
      4. Comment on reverse engineering rule code into DMN format - while this could be done via PRR-as-a-DMN-notation, its more likely that reversing rules into DMN will be a tool-specific function, not a “notation compliance” point…
      5. Fig21 had declarative rules executing in as-required order - this could be considered more rule programming than rule modelling, or a model requirement for PRR rules. Although a good idea we (industry) still need to “invent” a suitable “model” for such inferencing…
    6. DMN scope on metaphors *(RFP note)*
      1. May want to consider expanding from a DTable-only-for-version-1 to allow others (likely to be an RFP option anyway)
        1. Trees can be viewed as just “extensions of tables”, are easier to map to BPMN (and possibly from BPMN)
        2. If-then rules (c.f. PRR) which would allow DMN1.0 to potentially cover/address the SBVR-DMN-PRR-RIF continuity
      2. Ruleflows could be an option in the RFP, and some kind of subset / subclass of BPMN and/or allow import/export from/to BPMN process models (subject to the BPMN model containing decision-specific artifacts)
      3. DMN output to requirements could be…
        1. option to output to Excel in some format… but does DMN want a normative Excel / spreadsheet format?
        2. option to document the model in some way
        3. probably this requirement will be vendor-specific in terms of response…
    7. Certification and Compliance
      1. this is an understandable end-user requirement, but is not traditionally handled in the standard but as a set of compliance tests defined by OMG
      2. c.f. BPMN - are there compliance points for BPMN tools?
      3. *(RFP note)* Submissions should provide sufficient metadata to classify models effectively to allow rigorous compliance tests and compatibility checks with the specification…
  3. Other discussions
    1. Visual model detail in DMN models
      1. Model visual detail (in DMN) “at the same level as BPMN2 does”
      2. *(RFP note)* Goal should be to re-use appropriate BPMN artifacts, and ensure non-BPMN-relevant artifacts are distinct / unique
    2. Wiki actions from F2F1 are continued
  4. Planning
    1. RFP update (Paul) by mid July
    2. RFP conf call end July
    3. RFP draft submission to OMG by end Aug per 4-week-rule
    4. RFP discussion at OMG BMI at OMG Boston
    5. F2F3 at OMG Boston to gain + discuss feedback
      1. Start annotating RFP points with effects on submissions
    6. Possible discussions at World Decision Table Congress / BRForum and Rule Fest
    7. Revised RFP for submission at following OMG TM

RFP Conf Call meeting 9June2010

  1. Clarification of goals:
    1. modelling decisions as opposed to creating executable models
    2. no dependency on an execution context - that will be up to implementors to determine
      1. this also means any mapping to action semantics / PRR / BPEL is optional not mandatory, per BPMN model
    3. fulfil BPMN decision (business rule) activity
      1. although DMN will be aimed at being more generic than just BPM eg generic event-decision-action
  2. Actors:
    1. #1 = Business User, modeling decisions for some system that may be automated or not
      1. Events are: process step reached; situation event detected; etc
      2. Inputs are: business rules
    2. #2 = IT developer, automating the decisions defined by the business user
      1. Events are: associated service invocation with arguments
      2. Inputs are: modelled decisions from business user
      3. Per goals: this is outside the scope of DMN, but is a consideration
  3. Language for decision models
    1. should not be reinvented
    2. if not “SBVR fragments” then alternative needs to be justified
    3. should be re-usable in other decision models eg for models of PRR
  4. Classification / types of decision models
    1. Mandatory will be decision table
    2. Optional / future (/metamodel capable of extension for) other decision models
      1. inference rule models (which would be associated with PRR)
      2. decision trees
      3. statistical models (associated with PMML)
        1. example could be a score model: a subclass of a set of decision tables
      4. ruleflows / procedures tying together various decision submodels
  5. Supporting information
    1. DMN should fit in with methodologies and processes around decision modelling (as in BPMN)
      1. tracking and monitoring of decision useage / performance
      2. processes and workflows around developing and using decision models
  6. DMN RFP process actions
    1. F2F actions continued
    2. PV to prep presns for week prior OMG MN June10
    3. Wiki updates to be suggested to PT for initial pre-RFP comment
    4. next F2F July 5 London

F2F London meeting 4May2010

  1. Plan to publish RFP in Sept will be too early; however RFP work will continue for discussion in June
  2. Work on RFP draft as dmn_rfp_draft_09-12-xx.doc
  3. Next steps: updates to wiki on
    1. Jan - glossary
    2. Christian - delta with SBVR / decision modelling at IBM/Ilog
    3. Paul - decision model references

OMG Mar2010

OMG TM review of materials to date.

dmndevelopmentprogress.txt · Last modified: 2011/01/10 11:19 by pvincent