MidPoint 5 Vision

Last modified 25 Oct 2021 16:22 +02:00

MidPoint 5.0 is going to be a major release of midPoint. It is expected that it will bring important changes, especially in the foundations of midPoint platform, removing old technological debt.

Work in progress
This is still a rough plan. It can change on any moment.

Schedule

The schedule is not certain yet. We plan to have midPoint 4.4 as the LTS release in autumn 2021. MidPoint 5.0 may be the next LTS release, which might be released in autumn 2023. However, this plan may change any time.

Strategic Goals

  • Remove technological debt, give us more options, increase development velocity

    • Reduce dependencies on old Sun XML libraries.

      • This may mean moving away from JAXB completely, including the code-generation plugins

      • Moving away form XSD completely to Axiom

  • Stabilization for longevity and maintainability

    • Review/improve/stabilize Prism API

    • GUI API for GUI customization

Considered Goals

  • Separate end-user GUI

    • Tailored for self-service and simplified delegated management

    • Good UX, understandable to all users

    • User profile management

    • Credential management

    • Management of relations to other users/groups

    • Data protection: consent management, processing basis, SAR

    • Self-service registration

    • Interface to the federation: registration, etc.

    • Scalable for massive number of users

  • Archetype schema

    • Each archetype may bring its own schema extension

Design and Plans

Major Achievements

Progress towards midPoint 5.0 that we have achieved so far.

Prism and Schema Improvements

  • Move to Axiom

  • Remove <extension> where it is not necessary

  • Primitive type for definition/schema

    • Use in: resource schema, role schema, archetype schema

  • Get rid of the JAXB-induced dichotomies

    • PolyString vs PolyStringType

    • "plug in" created classes (PolyString) to the code generator. The generator will not create these classes, it will use them.

  • Avoid legacy (mutable) values, such as XMLGregorianCalendar and Duration. Use Java8 types instead (e.g. ZonedDateTime).

  • Dynamic schema

    • archetype schema, role parameters, resource connector configuration, shadow attributes

    • Problem with two parts:

    • Augmenting schema - Axiom/Prism side. How to use objects with extended schema.

      • We need an easy and elegant way how to apply schema definitions after parsing. Alternatively, plug in schema "providers" to the parsing process.

    • Schema lifecycle - How to deal with schema definition changes (e.g. adding mandatory item)

  • Support for "heterolists": collections of items of different types

    • Special item name for each type. Not supporting the <item xsi:type="…​"> method.

    • Can contain containers, will use container IDs for pats (e.g. …​/list/focus[123]/name)

  • Ordered items ("list")

    • Add delta for ordered items will always append. Not supporting inserting values inside the list or re-arranging the list. Replace delta must be used for better control over ordering.

    • We are not supporting sorted values, e.g. list of string that is always in alphabetical order. The client is responsible to maintain sorting.

  • Support for native "map" type, e.g. {en ⇒ foo, sk ⇒ fú, …​ }

  • 100% namespace-free operation

  • QName and URI are the same thing.

  • Move to explicit versioning (no version numbers in namespaces)

  • Prism definition from the top to the bottom

    • There will be prism definition for every item, including subitems in properties.

    • Property definition → Type definition. Type definition may be primitive or structured. Which means, that property definition may (formally) consist of other property definitions (e.g. orig). Even though these are formally property definitions, they cannot be used as properties (e.g. cannot be addressed by delta). They may be used as a source of mapping ($user/name/orig), but they cannot be used as a target of a mapping.

    • There will be primitive properties (e.g. String) and structured properties (e.g. PolyString).

    • Containers and properties are change boundary. Delta can be applied to container/property, but not to subitems in a property.

    • Axiom processors can be made to work directly with Prism definition. E.g. prism parser can product Prism definitions.

  • New code generator (Axiom-based)

  • Schema cleanup

    • resource/synchronization → resource/schemaHandling/objectType

See also Prism 5 Design Notes.

MidPoint Studio

Studio is a primary midPoint IDE now.

  • Studio needs to use Prism definitions instead of XSD for code completion. This should work in all data formats (XML, JSON, YAML). This will make switch to Axiom quite seamless.

  • We need support for Axiom editing: syntax highlight and code completion.

Decisions Made

  • Axiom as primary data modeling

  • Move to custom reporting (removing Jasper)

  • Move to IntelliJ-based MidPoint Studio

  • Special repository for PostgreSQL

  • ConnId2

  • Performance improvements: midScale

  • Data protection: midPrivacy

  • midScribe documentation generator

  • Data representation formats: XML → JSON/YAML

    • We still want to support XML, but more emphasis on JSON/YAML (defaults, documentation, etc.)

  • MidPoint Studio (IntelliJ-based)

  • Complex configuration in Studio

    • Move complex configurations/visualizations/wizards from web GUI to IDE

    • IDE as primary configuration tool

    • Keep identity administration in the web GUI

    • End users will use dedicated end-user GUI in the future

    • Expected som functionality overlap between end-user GUI / web / IDE

    • Three user groups: end users, identity administrators, configurators (deployment engineers)

Open Questions

  • Complex properties, e.g. do we want PolyString as container? Every structured data as container?

  • Matching rules: are we using them correctly? E.g. matching rules for polystring. Maybe they should be string matching rules. Or at least matching rules for primitive values.

  • Audit: how to store/process the data? MidPoint repo? Elastic?

  • What to do with resource wizard? Keep in GUI? Rework? Move to IDE?

  • Visualization? What to visualize? Where? GUI? Studio?

  • Improvement to:

  • UX?

  • External end-user GUI?

  • Explore the "next generation" tech, e.g. OpenTracing, new Spring extensions, etc.

  • Multivalue container IDs (keys), specificaiton in Axiom, etc.

Survey

Conduct a survey (early 2021) to help answer these questions:

  • What to do with Hibernate repository? Decommission? Keep?

  • Production-quality support for Windows platform?

  • Use Studio for complex configuration, git-based configuration management, less emphasis on Web configuration

Until We Get To 5.0

  • We have to live with both XSD and Axiom

Maintaining 5.0

  • We still need to support schema extensions in XSD. Capabilities of XSD extensions may be frozen at midPoint 4 functionality. All the built-in schemas will be Axiom.

Was this page helpful?
YES NO
Thanks for your feedback