SCM Meeting 2010-01-20

Agenda:

• Review new material on wiki
• Start listing common properties

Minutes:

Present: NickCrossley, PaulKomar, ScottBosworth, RobinFuller, SamitMehta, JeanMichelLemieux, PeterHack

Nick started with an overview of his suggestions for link information in service discovery. The intent is to allow a client to discover which links are provided by a given service, and by iteration over a set of services to find which services provide a given link. For example, the link from a work item to a change set might be provided by an OSLC Change Management service as a property on a work item or change request, or alternatively the same link might be provided by an OSLC SCM provider as a property on a change set, or both. Without this information, it is not possible for clients to provide things like traceability reports, or reports on defects fixed in a baseline, etc., without hard-coding information about a particular set of service providers. For example, without this information, a client might be written to work with ClearQuest and ClearCase, but that client might need to be modified or reconfigured to work with Change and Synergy, or with Bugzilla and Perforce (if the latter two became OSLC providers).

Nick acknowledged that the syntax of the proposed link discovery feature was dubious, and that work would be needed to make it technically correct, but the purpose of the initial draft was to allow the group to consider the principle.

Scott asked about links to two or more different providers of one given service - such as links between an SCM system to two or more CM systems, for example from tasks in Synergy to both change requests in Change and work items in RTC; Nick acknowledged that the proposal did not cover that. Scott also asked about whether or not the problem was already addressed by IanGreen's work on links - Nick said it was not. After the meeting, Nick talked to Ian and confirmed this.

We agreed that whatever capability was provided in this area should be part of a common spec to be implemented by all service providers, and not unique to SCM; Nick's proposal was drafted using OSCL namespace prefixes rather than SCM prefixes for exactly this reason.

Next, we went on to discuss common properties for SCM resources, filling out a resource definition page as we went. We started with baselines, copying many of the common properties used for change requests in the CM spec. We had some discussion about types of baseline, such as full vs. incremental, and extended the abstract data model to allow for incremental baselines, with a property that represented a link to the previous or full baseline.

Last, we discussed the properties for change sets, copying many of the same properties we had agreed for baselines. This led to a discussion of the cost of retrieving some link properties (relationships), such as the cost of finding all streams containing a change set in RTC. We agreed that it was OK for some of these queries to be slow, provided that the default behavior was not to retrieve all properties, but to retrieve just a default set.