AML Meeting June 2013 notes
Have to say something about what an implementation needs to do
- ADL round trip - ability to faithfully represent a designated subset of ADL
- Q: Is there any necessary serialization format? Should this be ADL?
- AML Core Zero - ADL subset
- Test is ability to represent and transfer between implementations / collaboratively define
- Another test ability to prove that data compatible with models produced by one system is consumable / valid with models produced by another system.
- AML Extension - superset, capabilities that go beyond ADL
- AML Core Zero - ADL subset
Will provide a library of test artifacts - regression test suite from ADL compiler.
- Need one ADL to XMI or related conversion to generate test cases
- Full compliance statement is target for the final submission
- Metamodel defined
- Format? UML?
- Closeness to ADL/AOM?
- UML Profile
- UML to DITA transformation exists (partial)
- Acceleo - MOF to Text generator
- Should metadata be a fourth profile?
- Should metadata be included as part of initial submission?
Need to differentiate metadata that is considered part of the artifact from that is a part of a registry. openEHR is based on ISO 13606 and CEN TC 251. CIMI is working on more extended version, will follow the lead of CIMI.
Issue of original language, contributors, keywords, author, etc. Preliminary decision is that metadata section will be a low priority by August deliverable. 13606 will be part of requirements for first submission and will be done time permitting.
Suggestion from Thomas Beale - Archetype as a package / Definition as Package / Terminology Bindings as Package.
Page 17 aom1.5.pdf
Question about ARCHETYPE model vs. BOUML
Best description of the archetype model is in the identifier document. RESOURCE_DESCRIPTION_ITEM is translatable area...
- If you don't know what your original language is, your tools aren't going to work.
Today's metadata is in the ADL 1.5 identifier document.
CIMI wiki page on ADL+1.5+Power+Syntax+Proposals
Will make a type called TERMINOLOGY_CODE which maps to TERMINOLOGY_CODE
Q: Do we need to represent both DIFFERENTIAL and FLATTENED archetypes in AML or can flattened be a representational form.
A: The difference between the two forms is differential_path in C_ATTRIBUTE.
"C_" means "Constraint for"
The root of an object is always a C_COMPLEX_OBJECT
The reason C_ATTRIBUTE has 0..n children is because you can end at C_ATTRIBUTE to remove something
C_ARCHETYPE_ROOT serves the role of external reference.
use_archetype is an external reference in ADL, which generates C_ARCHETYPE_ROOT in AOM.
Q: Do regular expressions parse into the various parts sections. A: see: 
CONSTRAINT_REF - reference is the at code and external reference is the URI
CARDINALITY defines how many things can go into the container
existence identifies mandatory, optional, prohibited
occurrences states how often the instances occur in data
OBJECT_GROUP - does not need to be modeled in initial CIMI release
Q: V statements - some of the V statements should be included in the initial release to show how they are done.
C_PRIMITIVE - will add terminology id qualified code.
Expressions - do we need to model the syntax of the expression in AML? The key is agreeing on the model paths. We have to agree on surface form to reference model elements.
Question on Paths - why are paths there instead of some "at" code - meaning of at code can be reused in multiple places in the model. The meaning of the code is the path - at0004 SBP can be used in 24 hour SBP vs. instantaneous SBP vs. target SBP
Alan James at Intermountian used dot-separated class names. Replacing dots with slashes.
Paths need to be recomputed when you save the model - if you edit the model and add an intermediate node, the paths need to be recomputed if you do.
at0002.0.1 - the dot notion represents "subsumed by", the meaning means redefined.
Need to capture requirements that the modeling tool won't break paths with insertion or deletion.