home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.umcs.maine.edu
/
2015-02-07.ftp.umcs.maine.edu.tar
/
ftp.umcs.maine.edu
/
pub
/
WISR
/
wisr6
/
proceedings
/
ascii
/
hobbs.ascii
< prev
next >
Wrap
Text File
|
1993-10-19
|
11KB
|
253 lines
A Uniform Data Model for Reuse Library Interoperability
Elizab eth T.Hobbs
ConQuest Software
9700 Patuxent Woods Drive, Suite 140
Columbia, MD 21046
Tel: (410) 290-7150
Email: hobbs@source.asset.com
Abstract
Existing reuse libraries use different data models, classification schemes, and terminology.
This diversity supports various domains, customers, and technology explorations. Unfortunately,
the differences inhibit sharing assets between libraries. The inability to share reusable assets
reduces the asset pool available to reusers, potentially causing redevelopment of assets. The
Reuse Library Interoperability Group (RIG) formed in 1991 to examine interoperability between
software libraries. One RIG technical subcommittee, TC2, has developed a data model defining
information about assets that libraries should exchange to support interoperability, known as
the Uniform Data Model (UDM).
Keywords: reuse libraries, interoperability, data models.
Workshop Goals: To share the UDM with workshop participants and receive feedback from
them.
Hobbs- 1
1 Background
The Reuse Library Interoperability Group (RIG) [1] formed in 1991. Terri Hobbs has chaired
TC2 since its creation in July 1991. TC2 took the task of developing a general data mo del to
support interoperability between many, widely divergent libraries. As part of this effort, TC2 first
developed a subset data model for interoperability, the Basic Interoperability Data Model (BIDM)
[2] that b ecame a RIG proposed standard. TC2 also completed the draft of the UDM. The UDM
includes the BIDM and expands upon it. In addition, TC2 has begun compiling a preliminary list
of vocabulary terms for use in some of the UDM (and BIDM) attributes.
2 Position
The UDM solves some of the difficulties with sharing assets between diverse reuse libraries. In
particular, it defines an important set of information that libraries should exchange ab out assets
to support interoperabilty.
2.1 Why Interoperability?
In [3], Tracz identified access to reusable assets as a conditionfor fostering widespread reuse. Access
involves locating reusable assets and having the ability to obtain them. Reusers working in one
library might have no knowledge of assets in other libraries. Even if they discover that an interesting
asset resides in another library,the reusers would still need accounts or other means of entering the
other reuse library. Interoperability, however, would allow libraries to make their publicly available
assets known to other libraries and would eliminate the need for reusers to have direct access to
other libraries.
Without interoperability, reusers browsing an unfamiliar library or receiving assets from an unfa-
miliar library would need to learn how the librarymo deled assets, how the classification scheme
worked, and how terms are defined. Reuse researchers and practitioners havelong recognized un-
derstanding as a crucial problem. Without interoperability, reusers would spend increased time to
understand an asset simply learning the other library's representation of that asset. With interop-
erability the other library's data model would map to the familiar library's model, eliminating one
potential source of understanding problems for the reuser.
2.2 Brief Introduction to the UDM
The UDM and BIDM define a meta-model based upon the methodology and concepts of ALOAF
[4 ] and CDIF [5 ]. The meta-model uses the following entities: classes, a class hierarchy, class
attributes, bi-directional relationships between classes, andrelationship attributes. Figure 1 shows
the classes, attributes, and relationships of the UDM.
BasedIGObject class provides the basis for the other classes in the UDM. The Assetclass contains
descriptive information about a reusable entity, or "asset" (with lowercase "a"). The Element
class models the discrete pieces of an asset, for example, documents, source code, and test cases.
The Library class provides information about repositories needed for the exchange of assets. The
Organization class describes entities like people, companies, and committees, defining only dat
Hobbs- 2
Figure 1: UDM Class Hierarchy
needed for the exchange of assets. During an asset transfer, the sending library must fill all
mandatory attributes and relationships. As an abbreviated example, if IBM authored a reusable
asset, Order_Tracking, in the application area, Manufacturing, with keyword Order, the appropriate
data would appear as follows:
Asset Organization
Domain = Manufacturing Address = Owego, NY
Keywords = Order Fax = (555) 555-5555
Name = Order_Tracking Name = IBM
WasCreatedBy ____-> Telephone = (555) 555-0000
The UDM and its subset, the BIDM, contain attributes with restricted vocabularies. As a sup-
plement to the UDM and BIDM, TC2 will produce a RIG document listing these vocabulary
terms. The attributes Domain, ConformanceToStandards, Format, and LibraryClassificationMech-
anism, for example, require controlled vocabularies. The preliminary vocabulary list for Domain
includes Expert_Systems, Configuration_Management, and Image_Processing, among many other
terms. The attribute vocabulary document will contain term descriptions so that mappings to
these terms remain consistent across libraries.
Hobbs- 3
2.3 What the UDM Accomplishes
The UDM gives libraries a common representation for sending assets to each other, eliminating the
need for every library to create mappings to all other libraries' data models. The UDM also iden-
tifies important characteristics for describing reusable assets, such as cost, certification, language,
warranties, target environment, location (IsLocatedIn), author (WasCreatedBy), and contents (Is-
MadeOf). In addition, the BIDM subset of the UDM defines a data model that has formed the basis
of the data model used by ASSET's (Asset Source for Software Engineering Technology) on-line
asset catalog. The UDM also provides a term set, consisting of the UDM attributes, relationships,
and attribute vocabularies, that libraries and people can use to associate theirdiverse terminology.
3 Comparison
The STARS (Software Technology for Adaptable, Reliable Systems) Asset Library Open Archi-
tecture Framework (ALOAF) project [4, 6] has examined many of the same issues as TC2 and
helped direct early TC2 efforts. The ALOAF includes a meta-model, aspects of which app ear in
the UDM meta-model. The ALOAF also defines the Common Data Model for interoperability. The
UDM uses some of the classes and attributes of the Common Data Model. The ALOAF, however,
also considers services for interoperability, which TC2 has just begun investigating. According to
Solderitsh [6], the ALOAF will incorporate the results of the RIG, bringing the two efforts into
agreement. Finally, the ALOAF project implemented and demonstrated interoperation between
STARS libraries.
Another effort, the ASSET-CARDS-DSRS (Central Archive for Reusable Defense Software) (De-
fense Software Repository System) project [7], also has investigated reuse library interoperability.
The project incrementally prototyped interoperability first between ASSET and CARDS, and then
among all three libraries. The implementation of interoperability among these libraries gave the
project valuable lessons learned, both related to technical and business issues. The lessons learned
included the need for a memorandum of understanding between the participating libraries, for an
interoperability plan, and for authentication of user distribution classes. TC2 made immediate use
of this effort by adding the DistributionStatementattribute to the BIDM.
The scope of both these projects and TC2 differ, making exact comparisons difficult. These two
projects have addressed many issues that fall outside the scop e of TC2. In the area of data
modeling, which these projects and TC2 both tackled, TC2 considereda wider range of libraries in
the development of the UDM than ALOAF and ASSET-CARDS-DSRS examined in the creation
of their data models.
References
[1] J. Moore, "Reuse library interoperability group (rig) executive board," in Proceedings of the
Ninth Annual Washington Ada Symposium, McLean, VA., July 1993.
[2] "Rig basic interoperability data model (bidm), rps-0001," tech. rep., Reuse Library Interoper-
ability Group, Apr. 1993.
[3] W. Tracz, "Reusability comes of age," IEEE Software, vol. 4, July 1987.
Hobbs- 4
[4] "Asset libraryop en architecture framework v1.2, contract no. f19628-88-d0031, publication no.
gr-07670-1317," tech. rep., prepared for the Electronic Systems Division, Air Force Systems
Command, USAF, by the Boeing Company, IBM Federal Sector Division, and UnisysDefense
Systems, Inc.,Aug. 1992.
[5] "Ase data interchange format - framework for modeling and extensibility, eia/is-81," tech. rep.,
July 1991.
[6] J. Solderitsch, "Making the case for interoperating reuse libraries," Nov. 1992.
[7] B. Fleming, "Asset-cards-dsrs interoperability: Lessons learned," in presentation to the RIG,
May1993.
4 Biography
Terri began her "official" involvement in reuse in 1988 when she joined the newly formed Westing-
house Reuse Working Group. She and another developer subsequently wrote the first version of
the Westinghouse reuse library mechanism, ReuSE. In 1992, Terri also examined the feasibility of
classifying software from source code automatically and developed a program to produce best-guess
classifications for data structure and numeric function Ada source co de. In addition,she has worked
to put reuse principles into practice while developing and maintaining different types of software
tools, written in FORTRAN, Ada, and C.Terri currently works for ConQuest Software, pursuing
her interest in text search and retrieval that began during the development of ReuSE.
Hobbs- 5