Support @base for resolvable and non-resolvable

[Sveino]

Summary

When converting between TriG / JSON-LD and CIMXML, the handling of base IRIs and UUID-based instance identifiers is currently inconsistent and causes ambiguity in identifier construction and serialization.
This ticket defines a clear and deterministic rule set for:

• Use of @base in TriG and JSON-LD

• Use of modelURL as dataset-level authority

• Serialization of UUID-based instance identifiers

• Differences between CIMXML and RDF syntaxes

The goal is to support object-oriented internal models (UUID without scheme) while producing portable, standards-compliant RDF serializations.

Current behaviour

During conversion from TriG → CIMXML, we observe dataset metadata such as:
<json-ld:base>http://www.Statnett.no/IGM/Nordic44_CGM#</json-ld:base>
with namespace:
xmlns:json-ld="https://www.w3.org/ns/json-ld#"

This appears inside a dcat:Dataset

At the same time:

• TriG and JSON-LD support @base

• CIMXML (not RDF/XML in general) currently does not support @base

• UUID-based identifiers are sometimes emitted as relative values

This issue is related to #9.

Technical clarifications (important)

CIMXML vs RDF/XML

• RDF/XML does support base IRIs via xml:base

• CIMXML (current implementation) does not support base IRIs

• Therefore, CIMXML must not rely on base resolution

rdf:ID vs rdf:about

• rdf:ID cannot start with a number (XML NCName restriction)

• rdf:about can start with a number, but then represents a relative IRI

• Relative IRIs require a base to become absolute

Design intent

We want:

1. Internal object identity to be simple and OO-friendly

→ UUID string without scheme (e.g. 26cc8d71-…)

2. External RDF identity to be explicit and deterministic

→ either HTTP-based or urn:uuid:-based

3. Identical local UUID tokens regardless of output format

4. No reliance on fragile URN base resolution rules

Decisions

1. Add modelURL

KGraph SHALL expose a dataset-level attribute:

modelURL : absolute HTTP(S) IRI

Example:
https://model4powersystem.eu/NO/

2. Rules for @base in TriG and JSON-LD

HTTP base

• modelURL SHALL be used as @base in TriG and JSON-LD by default

• modelURL SHALL end with /

@base <https://model4powersystem.eu/NO/> .
Rationale:

• Ensures predictable relative IRI expansion

• Avoids file-vs-directory ambiguity

URN base

• @base SHALL NOT be used to mint urn:uuid identifiers

• UUID URNs SHALL be serialized as absolute IRIs

Rationale:

• @base <urn:uuid:> and @base <urn:uuid:/> produce non-canonical or ambiguous results

• RFC3986 relative resolution with URNs is not portable across parsers

3. Rules for UUID-based identifiers

Internal representation:

mRID = "26cc8d71-12f1-4de9-9e68-125d95073a75"

Serialization:

Target	Output
HTTP model	https://model4powersystem.eu/NO/26cc8d71-...
UUID model	urn:uuid:26cc8d71-...

The choice is made at serialization time, not by changing the internal object model.

4. Rules for CIMXML output

• CIMXML SHALL NOT contain:

-- @base

• json-ld:base

• CIMXML SHALL always use absolute IRIs in rdf:about

• Relative identifiers SHALL NOT be emitted in CIMXML

Rationale:

• CIMXML currently has no base resolution mechanism

• Absolute IRIs avoid ambiguity and XML name constraints

Examples

TriG (HTTP-based model)

@base <https://model4powersystem.eu/NO/> .
@prefix cim: <https://cim.ucaiug.io/ns#> .

<26cc8d71-12f1-4de9-9e68-125d95073a75>
  a cim:PowerPlant ;
  cim:name "Melkøya" .

Expands to:
https://model4powersystem.eu/NO/26cc8d71-...

TriG (UUID-based model)

@prefix cim: <https://cim.ucaiug.io/ns#> .

<urn:uuid:26cc8d71-12f1-4de9-9e68-125d95073a75>
  a cim:PowerPlant ;
  cim:name "Melkøya" .

incorrect: TriG (UUID-based)

@prefix cim: <https://cim.ucaiug.io/ns#> .
@base <urn:uuid:> .

<26cc8d71-12f1-4de9-9e68-125d95073a75>
  a cim:PowerPlant ;
  cim:name "Melkøya" .