L8 Task Group Meeting

December 8 - 9, 2004

Minutes

Attendees:

Dan Gillman, BLS

Judith Newton, Ashton Computing & Management Services, LLC

Gail Wright, Oracle Corp.

Frank Farance, Farance Inc.

Jim Mercurio, VA

Paul Levine, Global e-Business Advisory Council, WG1

Kevin Keck, LBNL

Bruce Bargmeyer, LBNL

Topics:

WG2 Meeting in Washington Report

Frank Farance reported:

Thanked Dan and BLS for hosting this meeting.
19763 Metamodel Framework – some progress, but don’t know what problem its solving. Difficult topic.
Discussed 20944. Some parts in ready state, some not. Moved ones not ready back to state 0. Other parts in CD status.
19773 Some parts expired. Doug revived them administratively. 13 parts now in total.
20943 Part 2 (XML part) withdrawn temporarily due to no progress.
Common Logic – good, informative presentation
Not discussed: 24706 and 14957
Participation of US committee was good. Some free form discussion among US participants about technical issues – area for improvement.

Dan Gillman reported:

Thought event went well
Main surprise – discussion of XMDR – activities are happening outside of the US tag; however, got wrong impression that this had sole US involvement.
Ray posted WG2 meeting documents. Find on metadata-standards.org website. Click on Library link. Scroll down to Meeting Documents. Should be the first in list: Washington Meeting. Also, the WG2 Resolutions and Meeting Report should be on the SC32 website; however, it is not currently posted.
The death date for 11179-3 is coming up (2 years with no change or reissue).

WG1 Meeting in Tallinn Estonia Report

Paul Levine reported:

Paul presented the WG1 Report as he provided to the L8 reflector mailing list.
WG1 meeting was hosted by the Finland national body. Held on October 25-29, 2004.
Agenda: Process contributions on ISO/IEC 15944 Parts 2-5 in preparation for FCD ballot on Part 2 and 2nd CD ballots on Parts 3-5
Came up with a revision document for Parts 2-5 (WG1 N269-N272) which consolidated resolution of ballot comments and editing instructions.
Discussed issues with having continued Microsoft support for this activity.
Discussion on patent rights. Frank – can’t make changes to patent language in committee. Paul – lots of discussion on intellectual property rights. Why have patent language in ISO documents? The issue is that the statement does not state what the patent rights are (if any) and can be misleading as the standard is created independently of the creation of patents. However, parties might make claims for patent infringement. In creating the standard, we could infringe on patent rights without knowing it. Our non-legal interpretation is that ISO includes this language to protect itself against any pre-existing patents. Dan – the purpose of this discussion is to figure out what the US tag’s position is. Frank – the secretariat made this patent disclaimer in conjunction with attorneys and believe it has some importance; therefore. Committees have no say in this matter. From the US point of view, we can say that the patent right policy discussion is “out of order” to discuss in WG1. We cannot clarify or state what the policy is. All agreed.
Frank would like to create a framework for metadata service standards on the web (vs. web services) that includes usability, accessibility, privacy policies, etc. Frank can approach a company/firm that does this type of work to get input. Dan believes that there is an enormous amount of work already done in this area.

XMDR

Bruce Bargmeyer and Kevin Keck reported:

Website: www.xmdr.org
Looking at persistent stores (different management systems: Relational, Object, XML, RDF, etc.) and analyze the query support that each technology provides. For ontologies, query technologies may be insufficient in terms of performance and levels of granularity.
Project Scope: 11179 Part 2. Start with existing classification schemes (SNOMED, LOINC, NAICS, etc.) and use them to classify Administered Items. Determine the best models/technologies for such classification schemes that support the types of queries required – including ontology/semantic web type structures and queries. Will likely not choose a relational or object management system. Part 3 could still be implemented in a relational system as part of a hybrid solution with additional integration work. Looking at Tucana and Protégé and ontology editors. Plan to post recommendations on the website in January 2005. If RDL-based, then must support RDQL or SPARQL. Have a bias towards open source solutions.
Content of current metadata registries is useful to semantic web in that they provide “agreed upon” terminology. There is more work in relating classification schemes to data elements. In one case, a classification scheme item can categorize the data element. In other case, the classification scheme may be the value domain of the data element. “Concept systems” is synonymous with the “Classification scheme”. Could extend usage of 11179 registries to be the semantic servers on semantic webs.
Referenced Sam Chance’s presentation. Kevin - want metadata registry to be more “semantics aware” using software accessible descriptions that describes relationships (for example: between Administered Items). For example: A particular concept “is comparable” to another concept because they have a common ancestor. Frank – This is a problem with software systems as they cannot interpret an “is a” relationship. Such relationships are better interpreted by humans. Kevin - Ontology: A model that software uses to describe a data set semantically vs. structurally.
Frank – Going back to the Terminology Server idea – there is a relationship between concepts and signs, manage relationships among themselves, etc. If using object oriented technology “class” and 11179 is a sub-class with specialization which is apparent in the attributes we use. 11179 specializes in the “data industry” which is different than specializing in “terminology/linguistics”. There can be overlap, but there are also distinctions. Bruce agrees. However, the metadata registry may be reused for secondary purposes with different intent than originally envisioned.
OMG working on ODM (Ontology Definition Metamodel).

Dimensionality

Dan - In WG2, agreed upon a definition, but don’t have consensus on how this fits in the model. Could not find WG2 consensus definition; however, did find: Dimensionality - semantically similar units of measure. Dan believes dimensionality should not be part of the Conceptual Domain. Have attached it there because Dimensionality is the conceptual counterpart to Units of Measure, which is representational.
Dan - Units of measure, dimensionality, data type, etc. seem to be related, yet are independent of each other. Are units of measure only associated with non-enumerated domains? Think of unitless data like a ratio – but do need to know the units of both the numerator and the denominator.
Frank – Enumerated vs. Non-enumerated – don’t believe this is a factor or makes a difference for units of measure. Ratio – algebraic model - applies in some circumstances, but not in others. Original question – why do we care? Would like to have unit of measure at the Conceptual Domain level (e.g. Temperature in Degrees Fahrenheit). Frank Olken joined and disagreed, saying that “Degrees Farenheit” is definitely representational.
Lots of discussion on the subject. Looked at 20943-1 example. No consensus. To be continued.

20944

Frank Farance reported

ISO/IEC 20944 set of standards addresses the need to access the contents of an ISO/IEC 11179 metadata registry by computer. 20944 provides the provisions necessary to describe how to achieve this. It does not provide implementation details.
Each new technology, e.g., programming languages, needs a binding. A binding is a map from a standard or framework to another.
The 20944 standard is broken into pieces along the lines of conformity. General aspects are separated from (language) specific requirements. This is based on good design principles.
Bindings are provided in 3 types: codings, API's, and protocols. Respectively, they describe data transfer rules, control transfer rules, or both. Data transfer handles situations such as mapping the contents of a registry to an XML schema. There are 2 kinds of control transfer: 1 way and 2 way. An example of 1 way control transfer is the use of "go to" statements in a programming language. Typically, the "go to" tells a program where to jump to, but there is no way to return. An example of 2 way control transfer is a function or subroutine call. The program jumps to another section of the code, and returns to the point of the jump when the function or subroutine is finished.
Frank recommends going to the example in clause 7.1 of 20944-01. Here, read the other parts as required to understand the sequence of steps (e.g., read parts 01, 41, 40, 03, and 81).
Kevin Keck was worried about the scalability of the proposed architecture in 20944. He proposed a stateless, rather then state-based, authentication process. Frank said it is too complex to encapsulate all authentication needs of every user at the enterprise level. Instead, 20944 provides a hook to incorporate any authentication required by an application. In a sense, agreement upon a specific authentication technique is outside the scope of 20944. Furthermore, Frank stated that it is possible to implement a stateless system with the current specification.
Frank stated that the design criterion he aims to achieve is the support of implementation varieties that are interoperable, yet can choose their own implementation strategy (variety control defined in ISO/IEC Guide 2).

19773

Frank Farance reported

Read documents by working backward. Start in Part 8, for instance. The best arrangement for the parts is open for debate.
The important technical detail is the complex datatypes defined: reflit, multi-string, multi-text, multi-value, and multi-data. All these are based on the ISO/IEC 11404 "choice" datatype generator. Parts 3 and 4 give the details for these datatypes.

11179-3 3^rd ed

Bruce Bargmeyer and Kevin Keck reported on Issue 136

Bruce and Kevin discussed their proposal to modify the "recursive" relationships on the following classes in the model: Classification Scheme Item; Value Domain; Conceptual Domain; and Data Element Concept.
The problems are that the current presentation in the UML diagram does not specify directionality and it separates the "recursive" relationships from the concepts. The "recursive" relationships on a concept are part of what it is and need to be recorded that way.
The proposal modifies the UML diagram to relate the "recursive" relationships in a "part of" relation on the class. See the Issue 136 contained in the SD14 standing document, obtained from the WG2 web site (http://metadata-standards.org) document library in the Washington meeting listing.
The participants at the meeting were in favor of the proposal. Dan suggested there may be problems with extending the structure to value domains, but he was in favor of the structure on the other classes.
This led to further discussion of value domains and the problem of domains of mixed type. That is, there are examples that have both enumerated and non-enumerated components. The proposed solution is to create 3 value domains: one for the enumerated values; one for the non-enumerated values; and one as a parent value domain that contains the other 2. The participants all felt this solution worked.
This led to further discussion of dimensionality, which is Issue 124 in the SD14. The participants agreed that the definition as it now reads is correct. The algebraic formulation of dimensionality is incomplete. Characterizing operations are necessary to completely define the equivalence classes. Kevin and Frank Olken were worried about this, but Kevin seemed satisfied after the discussion. Frank was part of an equivalent discussion in the Washington WG2 meeting and was similarly satisfied.