Mapping Notes
Notes from Mapping review 2021/2022.
Challenges
-
Complexity
-
Difficult to understand mappings
-
Solution: better documentation? docs? Dedicated chapter in book?
-
Solution: Metaphors? Describe mappings using terms that common people/engineers understand?
-
Solution: More pictures, diagram, visualizations
-
-
Difficult to configure mappings
-
Solution: more reasonable defaults? (e.g. default range)
-
-
Terminology
-
Weak,normal,strong: better names?
-
-
Apparent paradoxes
-
Condition vs range
-
Null, non-null
-
-
-
Lightweight synchronization (recon)
-
Avoid reading data from resources when not needed (even for strong mappings)
-
Solution: "incremental" (fully relativistic) mode?
-
-
Single-resource inbound sync (in multi-resource configurations, with multi-source configuration, overlapping mappings)
-
Solution: metadata + smart range?
-
Solution: caching source data in shadow?
-
-
Consideration: can we evaluate mappings only for one resource? E.g. only for HR?
Problem: Evaluate inbound mappings, store data in user, but the deltas are lost. If we would continue, we can use the deltas to relativistically propagate changes to outbound mappings, maybe even skipping reading current values for strong mappings (see above).
Similarly, evaluating only for one outbound resource can be problematic. Where do we get the deltas from?
-
-
Implementation
-
Code could be improved, made nicer and more consistent. TODO
-
Containers as mapping output, e.g. to parse complex inbound attibutes presented as string, storing output in extension container.
-
Possible Optimizations
-
Skip reading current values if we have only strong mappings to single-valued attributes. Or maybe have "always overwrite" mode. May be faster than read-check-update.
-
Mark expensive mappings, re-evaluate them only if really necessary.
Requirements
-
Roll-out new provenance metadata to production. We need to record mapping that produced value to have smart range expressions.
-
New scenarios needed by "flexible correlation"
-
pre-processing of resource object being synchronized (before correlation or even before classification)
-
interacting with external correlator (i.e. transformation between resource or midPoint schema and correlator schema)
-
What To Do
Documentation
-
Reference documentation
-
Describe algorithms:
-
Domain/range, how condition is applied, merging values from several mappings (consolidation), relavive/absolute mode, etc.
-
-
Other subpages
-
Start with full complexity, describe all options
-
Guide/tutorial
-
Starting from simple, explaining how to do things.
-
where to place it?
-
Later make a dedicated book chapter from the text? Maybe we would need to add object templates and roles to the mix?
-
Examples: how to set up mappings for authoritative source, how to set up mappings with two sources, etc.
-
-
Implementation documentation
-
Document the theory?