Estimation and Measurement Telecon, 2009-07-24

Attendees

AndrewCanham, AndyBerner, ArthurRyman, LeeFischman, MurrayCantor, ScottBosworth

Regrets

LawrencePutnamJr, SteveAbrams

Minutes

1. Re-estimation use case - AndyBerner

AndyBerner - We completed the discussion of this use case. No additional information to discuss at time.

2. Review working draft of EMS 1.0 Part 0: Primer - ArthurRyman

Updates from review at last telecon

ArthurRyman reviewed the updates.

ScottBosworth - the architecture paragraph looks good.

Update on OSLC leads discussion on use of RDF/XML

ArthurRyman reviewed the discussion on how to use RDF/XML in the context of REST. Some key points are:

• RDF is an appropriate way to model Web information and to specify the semantics of queries (since the query needs to refer to the semantics of the underlying data model)
• However, RDF uses a very fine-grain notion of resources.
• Not all RDF resources have dereferencible URIs, e.g. you'd get a 404 if you tried to GET them.
• URIs are used to identify resources which may or may not be retreivable via GET.
• In our specifications, only the course-grained resources shouls have a REST API.
• RDF/XML can be used as a representation of a resource, e.g. when doing a GET of a plain URI (no query parameters). We can specify concrete RDF/XML formats as representations for our resources.
• RDF/XML can also be used to describe information about resources. This is a subtle but very important distinction. For example, a GETof a URI with query parameters should be regarded as requesting a description of the resource identified by the plain URI, possibly including information about related resources. The plain URI + query parameters is a different URI whose representations are only related to those of the plain URI.
• RDF allows you to describe collection/container resources, but these can sometimes be avoided by using multi-value predicates instead, e.g. a Project has zero or more Estimate resources, as opposed to having a single EstimateContainer resource that contains the actual Estimate resources. Only use explicit containers when there are additional semantics, e.g. the members are alternates, the order is important, etc.
• Resource creation can be handled by POSTing a resource representation to a factory resource, e.g. EstimateFactory, assocated with an RDF type, e.g. Estimate. The factory is the container of all resources it created (refered to in RDF as the extension of the type) and can be queried, e.g. to get subsets of its resources, e.g. the Estimate resources associated to a given Project resource.
• RDF/XML uses a "striped" style of XML where the alternation elements identify predicates that relate subjects and objects. The object of a top-level predicate can be used as the subject of a nested predicate, and so on. RDF/XML allows the object of a predicate to be a value, which may be an arbitrarily complex fragment of XML. These are called XMLLiteral datatypes.
• In our spec, things like probability distributions should be modelled as values, not resources. AndyBerner - Can our QuantileFunction described the situation where I specify the probability of schedule completion at say, 25%, 40%, 70%, 90% - where this is an arbitrary list of probabilities? ArthurRyman - In a QuantileFunction the probabilities are equally spaced, so in your example the greatest common divisor is 5%. You therefore specify numberOfQuantiles = 20 and ignore the values you don't care about. AndyBerner - Is that acceptable to estimate vendors? LeeFischman - Yes, that's what we do. AndrewCanham- Yes.

Review outline of Reestimating scenario

ArthurRyman reviewed the new scenario in the Primer. There was agreement on the overall content. The XML examples will reflect the current thinking around RDF/XML.

[ACTION] ArthurRyman to complete the XML for the Reestimation scenario in the Primer.

3. Other Business - All

AndyBerner - Next week I'd like to discuss how to represent "effort by date".

ArthurRyman - Please post a proposal in advance to the workgroup can have a productive discussion.

[ACTION] AndyBerner to post a proposal on how to represent effort by data.