Estimation and Measurement Telecon, 2010-04-30

See Weekly Meeting Logistics for telecon information.

Previous telecon 2010-04-23

Next telecon 2010-05-07

Attendees

AndrewCanham, AndyBerner, ArthurRyman, LawrenceMandel, LeeFischman

Minutes

1. Implementation Status - All

No implementation progress reported.

2. Review of REST API Data Models - ArthurRyman

Metric Entities

Work Breakdown Structures

LeeFischman: Why is just MS Project defined? Customers want to use the native WBS provided by their project management systems, e.g. Primavera.

ArthurRyman: You can use any format. You need to define a URI to identify the format. We can include it as a standard URI under ems:WbsFormat. We do not want to treat any vendor preferentially. However, most vendors appear to support MS Project for interchange. Please propose another format if you think it would be useful.

LeeFischman: Can ems:Map resources be included in a WBS?

ArthurRyman: The spec treats WBS are an opaque document. However, MS Project XML does allow the inclusion of extension elements so we could define a standard for include ems:Map in MS Project.

Probability Distributions

AndrewCanham: Can a measure distribution contain more than one format of probability distribution?

ArthurRyman: No, there is one per measure distribution. However an estimate or scenario could provide multiple assumptions and predictions for the same metric and those could use different probability distributions.

AndrewCanham: So if a consumer wanted to use a non-standard probability distribution then they could provide that plus a standard one?

ArthurRyman: Yes. That way any other consumer could fall back to the standard one if they didn't understand the non-standard one.

AndyBerner: Will the service be able to convert from one probability distribution to another? For example, suppose a client wants percentiles but the estimate is in another format?

AndrewCanham: That type of function should be the responsibility of the client.

AndyBerner: Do we need the quantile function? We can represent it using the cumulative distribution function.

ArthurRyman: We could eliminate all the others since cumulative distribution function can approximate any of the to whatever accuracely is needed. However, it's useful to have more specialized distributions like quantile functions since they are frequently used.

AndrewCanham: The current list is fairly simple to implement and shouldn't be a problem.

Vote

The workgroup voted to make the Data Models a Draft Specification.

3. Standard URI Status - ArthurRyman

ACTION: ArthurRyman will complete the Standard URIs document and review it in the 2010-05-07 telecon.

4. OSLC Core - ArthurRyman

LawrenceMandel: Do we need to specify the use of the query parameter syntax?

ArthurRyman: The EMS 1.0 spec does rely on a simple search mechanism, e.g. find all scenarios for a given project. At a minimum the spec requires support for oslc.where in order to perform searches on the domain entity lists. The other query parameters make it easier for consumers to use the EMS 1.0 service. We could make those query parameters optional if implementation experience reveals they are expensive to implement.

Comments

Add your comments here: