[OASIS Issue Tracker] (OSLCCORE-39) Names and descriptions of impact analysis direction properties are misleading

[OASIS Issues Tracker <workgroup_mailer@lists.oasis-open.org>]

    [ https://issues.oasis-open.org/browse/OSLCCORE-39?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=61645#comment-61645 ] 

James Amsden commented on OSLCCORE-39:
--------------------------------------

The typical meaning of a dependency in programming languages (package import), UML, SysML, common use, etc. is that:
1. dependencies are directional: source -dependency-> target
2. changes in the target of a dependency (e.g., an imported package) may have an impact on the source, but generally changes in the source have no impact on the target. 
3. The implications of the above points are important in change and impact analysis as they influence the work that has to be done to assess change in a system.

RDF properties define a semantic relationship between a subject and object as an asserted triple {subject predicate object}, mathematical relation predicate(subject, object) or function predicate:subject --> object. This semantic relationship might be considered a dependency, and could establish a default direction for impact analysis (subject depends on object). 

However, in common practice, RDF implementations commingle semantic relationships with persistence, often designing properties to support assertions stored with the subject. So there is in practice no clear relationship between RDF properties and dependency relationships for impact analysis. 

This is why OSLC introduced the impact vocabulary. This vocabulary essentially annotates an rdf:Property with implications for impact analysis:
1. impact is in the direction of the rdf:Property
2. impact is in the opposite direction of the rdf:Property
3. impact is in both directions (whether the property has an inverse or not)
4. the rdf:Property does not represent a dependency and there is no impact. It simply supports assertion of knowledge.

Regarding Nick's proposal, we have agreed that:
1. Either this impact vocabulary should be normative in OSLC Core 3.0 or it should be removed and perhaps added as a committee note. That is, it is not clear what non-normative vocabulary would mean in a specification.

2. Regardless of the resolution above, we will need to retain the existing impact vocabulary terms for compatibility, but can mark them as   vs:term_status "archaic" (see http://www.w3.org/2003/06/sw-vocab-status/note).

> Names and descriptions of impact analysis direction properties are misleading
> -----------------------------------------------------------------------------
>
>                 Key: OSLCCORE-39
>                 URL: https://issues.oasis-open.org/browse/OSLCCORE-39
>             Project: OASIS OSLC Lifecycle Integration Core (OSLC Core) TC
>          Issue Type: Bug
>            Reporter: Nick Crossley
>            Assignee: Nick Crossley
>            Priority: Minor
>
> The OSLC Core 3.0 vocabulary includes the terms defined in the Vocabulary Annotation Vocabulary (an addendum at http://open-services.net/wiki/core/Vocabulary-Annotation-Vocabulary/) from OSLC 2.0 days - though these terms were not published as part of OSLC 2.0 Core.
> However, two of the terms proposed for impact analysis have misleading names, and very misleading descriptions in the proposed Core 3.0 vocabulary:
> <http://open-services.net/ns/core#UpstreamImpact>
> 	rdfs:comment "Subject resources, indicated by triples of the subject property, have an upstream impact from the object resource." ;
> Impact is always felt downstream, as per the definition of these terms as described at http://open-services.net/wiki/core/Vocabulary-Annotation-Vocabulary/. What this term is trying to describe is that the link is pointing upstream, and so impact analysis should consider impact to flow in the opposite direction of the link. For this reason, both the original name and the 3.0 description are misleading.



--
This message was sent by Atlassian JIRA
(v6.2.2#6258)