Computer Standards & Interfaces 12 (1991) 185-192 185 North-Holland

Database management standards: status and applicability

Leonard Gallagher U.'; National Institute of Standards and Technology, Gaithersburg, MD 20899, LiSA

A b.s'tract

Gallager, L., Database management standards: status and applicability, Computer Standards & Interfaces 12 (1991) 185 192.

This article presents the current status of existing and emerging ISO, ANSI, and FIPS standards for database management, specifically Database Languages NDL and SQL and Remote Database Access (RDA). It describes the general content of each standard and discusses its applicability, availability, completeness, maturity, stability, existing usage, and known limitations. Where appropriate, it also addresses the availability of conformance test suites and future plans for enhancements and follow-on standardization efforts.

Kev~vordw ANSI: database; data model: FIPS; ISO; network: NDL: RDA; relational: standard: SQL.

1. Introduction

Database management standards have been in existence since 1986 with parallel publication of the Database Language NDL standard for net- work model applications and the Database Lan- guage SQL standard for relational applications. These original standards provide a data definition language (DDL) and a data manipulation lan- guage (DML) for creating and accessing data structures of the underlying data model in a single data processing environment. Later revisions pro- vide interfaces to multiple programming lan- guages, interactive users, and remote sites. In par- ticular, the emerging Remote Database Access (RDA) standard will provide database interoper- ability in an open systems environment.

This article presents the current status of ISO, ANSI. and FIPS adoption of the NDL, SQL, and RDA standards as well as the status of emerging revisions and follow-on efforts. It describes the general content of each standard and discusses its applicability, availability, completeness, maturity, stability, existing usage, and known limitations. Where appropriate, it also addresses the availabil- ity of conformance test suites and future plans for enhancements and follow-on standardization ef- forts. A similar analysis of other important stan- dards for application portability and interoper- ability (e.g. GOSIP, POSIX, PHIGS, IRDS, ODA,

etc.) is contained in a new NIST publication. Application Portability Profile." The U.S. Govern- ment~ Open System Environment Profile OSE/1 version 1.0, available from NTIS as NIST Special Publication 500-187.

Sections 2, 3, and 4 address the SQL, NDL, and RDA standards respectively. Section 5 states some conclusions and Section 6 presents the full names and titles of organizations and standards groups mentioned in this article.

2. Database language SQL

Existing specifications:

Emerging specifications:

Sponsoring organizations:

1SO 9075 : 1989

ANSI X3.135-1989 ANSI X3.168-1989

FIPS 127-1 February 2. 1990

ISO/ IEC DIS 9075: 199x, SQL Re- vision ANSI dpANS X3.135-199x, SQL Revision (Expected 1992)

I S O / I E C J T C I , ANSI X3, and NIST

0920-5489/91/$03.50 ,~', 1991 - Elsevier Science Publishers B.V. All rights reserved

186 L. Gallagher

Description Database Language SQL specifies data defini-

tion, data manipulation, integrity checking, and other associated facilities of the relational data model. In addition, SQL specifies components that support access control, programming language in- terface, and data administration. The basic struc- ture of the relational model is a table, consisting of rows and columns. Data definition includes declaring the name of each table to be included in a database, the names and data types of all col- umns of each table, constraints on the values in and among columns, and the granting of table manipulation privileges to prospective users. Ta- bles can be accessed by inserting new rows, delet- ing or updating existing rows, or selecting rows that satisfy a given search condition for output. Tables can be manipulated to produce new tables by cartesian products, unions, intersections, joins on matching columns, or projections on given columns.

i

SQL data manipulation operations may be in- voked through a cursor or through a general query specification. The language includes all arithmetic operations, predicates for comparison and string matching, universal and existential quantifiers, summary operations for max /min or count /sum, and GROUP BY and HAVING clause to parti- tion tables by groups. Transaction management is achieved through COMMIT and ROLLBACK statements.

The standard provides language facilities for defining application specific views of the data. Each view is the specification of database oper- ations that would produce a desired table. The viewed table is then materialized at application execution time.

The SQL standard provides a Module Lan- guage for interface to other languages. Each SQL statement may be packaged as a procedure that can be called and have parameters passed to it from an external language. A cursor mechanism provides row-at-a-time access from languages that can only handle one row of a table at one time. Access control is provided by G R A N T and RE- VOKE statements. Each prospective user must be explicitly granted the privilege to access a specific table or view using a specific statement.

The optional SQL Integrity Enhancement facil- ity offers additional tools for referential integrity, C H E C K constraint clauses, and D E F A U L T

clauses. Referential integrity allows specification of primary and foreign keys with the requirement that no foreign key row may be inserted or up- dated unless a matching primary key row exists. Check clauses allow specification of intercolumn constraints to be maintained by the database sys- tem. Default clauses provide optional default val- ues for missing data.

An Embedded SQL specification (ANSI X3.168) provides an SQL interface to program- ming languages, specifically Ada, C, COBOL, FORTRAN, Pascal, and P L / I , by specifying the effect of syntactically embedding SQL statements into otherwise conforming application programs. Applications may thereby integrate program con- trol structures with SQL data manipulation capa- bilities. The Embedded SQL syntax is just a shorthand for an explicit SQL Module accessed from a standard conforming programming lan- guage.

Applicability The purpose of a database language standard is

to provide portability and interoperability of database definitions and database application pro- grams among conforming implementat ions. Database Language SQL is appropriate for all database applications where data will be shared with other applications, where the life of the appli- cation is longer than the life of current equipment, or where the application is to be understood and maintained by programmers other than the origi- nal ones. The SQL standard is particularly ap- propriate for database applications requiring flexi- bility in the data structures and access paths of the database. It is desirable both for applications un- der production control and when there is a sub- stantial need for ad hoc data manipulation.

Use of the SQL standard is appropriate in all cases where there is to be some interchange of database information between systems. The schema definition language may be used to inter- change database definitions and application specific views. The data manipulation language provides the data operations that make it possible to interchange complete application programs. Used with the RDA remote database access stan- dard or with a generic data interchange standard such as ASN.1, data occurrences may also be transferred in a standard manner.

Database management standard~" 187

Le~el of consensus The ANSI, ISO, and FIPS specifications are

essentially identical. ISO has not yet formally adopted Embedded SQL (ANSI X3.168) but it is forthcoming in the next ISO version (DIS 9075: 199x). FIPS requires ANSI Level 2 confor- mance, with programming language bindings to one or more FIPS programming languages, and requires a FIPS Flagger to flag extensions. SQL has been adopted as the database management component of X/Open , OSF, SQL Access, and other vendor consortia.

Product availabili(v Numerous implementations of the existing SQL

standard exist on all levels and brands of equipment. Vendors are vigorously implementing additional features from the proposed 1992 revi- sion. An SQL validated processor list is available from NIST (see Conformance testing below).

Completeness The existing SQL standard specifies data defi-

nition, data manipulation, access control, and limited integrity constraints. The emerging, up- ward compatible revision expected in 1992 will include substantial additional features for schema manipulation, dynamic SQL, exception handling, enhanced integrity constraints, transaction mana- gement, and data administration.

Maturity The SQL data model is based on the relational

model first published in 1969. The first commer- cial systems were available in 1979 and the first SQL standard was published in 1986. The basic elements of the language are very mature and will not change.

Stabili(v The SQL language has firm mathematical

foundations in the first order predicate calculus. Standards groups and vendors are firmly com- mitted to upward compatibility in revisions and future extensions. Existing features are expected to remain stable for the foreseeable future.

De facto usage The SQL language is the de facto industry

standard for interface to relational database management systems. Implementa t ions are

ubiquitous and are available on all sizes and types of information processing equipment. The existing standard has been used successfully in many pro- curements.

Limitations The existing standard is specified for standa-

lone, single environment databases. Specifications for access to remote heterogeneous sites are under development in an emerging ISO Remote Data- base Access (RDA) specification (see RDA sec- tion). Specifications for distributed database management are in the very early stages of devel- opment. Tools for the support of object-oriented data management such as triggers, assertions, user defined datatypes, domain and table hierarchies, and stored procedures are under active considera- tion as follow-on enhancements to the SQL stan- dard.

Conformance testing Version 2.0 of the NIST SQL test suite has

been publicly available since December, 1989, and a formal SQL test service was instituted by N1ST in April, 1990. The SQL test suite measures con- formance to both required and optional features of FIPS 127-1.

Future plans An emerging SQL2 specification, with features

identified in the Completeness paragraph above, is expected to be adopted by ANSI and ISO in 1992. Further enhancements, including those identified in the Limitations paragraph above, are expected in the 1995 time frame. The following two para- graphs give a more detailed description of emerg- ing SQL2 and SQL3 specifications.

Emerging SQL2 specification A substantial upward compatible enhancement

of the existing SQL standard, often called SQL2, has already been specified by the ANSI and ISO SQL development committees. It will standardize a number of SQL features not included in the original specification because they were not com- monly available in SQL products. The technical specification for SQL2 is quite stable; only a very few features are controversial and subject to mod- ification. SQL2 is intended to be a superset of SQL-1989 that replaces the existing SQL standard. Review copies of SQL-199x are available after

188 L. Gallagher

May 1991 from GLOBAL Engineering (reference dpANS X3.135-199x) or from OMNICOM (refer- ence ISO/ IEC DIS 9075 : 199x, SC21 N5739).

SQL2 significantly increases the size of the SQL language to include a schema manipulation language for modifying or altering schemas, schema information tables to make schema defini- tions accessable to users, new facilities for dy- namic creation of SQL statements, and new data types and domains. Other new SQL2 features in- clude outer join, cascade update and delete refer- ential actions, set algebra on tables, transaction consistency levels, scrolled cursors, deferred con- straint checking, and greatly expanded exception reporting. SQL2 also removes a number of restric- tions in order to make the language more flexible and orthogonal.

SQL2 was registered as an ISO/ IEC Draft International Standard (DIS) and as an ANSI dpANS in early 1991 and is currently undergoing an ISO/ IEC national body ballot for final adop- tion as an international standard. Parallel processing to adopt the identical specification as an American National Standard is taking place in the United States. Adoption as a revised FIPS PUB 127 is expected in calendar year 1992.

Emerging SQL3 specification A second substantial SQL enhancement, often

called SQL3, is under development by ANSI and ISO SQL specification committees, with publica- tion expected in the 1995 time frame. SQL3 is a forward looking SQL enhancement that intends to provide additional facilities for managing object- oriented data and for forming the basis of 'intelli- gent' database management systems. It includes generalization and specialization hierarchies, mul- tiple inheritance, user defined data types, triggers and assertions, support for knowledge based sys- tems, recursive query expressions, additional data administration tools, standardized database ex- po r t / impor t facilities, and progress toward dis- tributed database management. These features are preliminary and subject to significant modifica- tion and improvement before final adoption. An early draft of these specifications should be availa- ble from OMNICOM after January 1992 as a WorKing Draft SQL Revision.

Alternative specifications: None.

3. Database language NDL

Existing specifications:

Emerging specifications: Sponsoring organizations:

ISO 8907 : 1987

ANSI X3.133-1986

FIPS 126 March 10, 1987

None ISO/ IEC JTC1, ANSI X3, and NIST

Description Database Language N D L specifies syntax and

semantics for data definition, data manipulation, and other associated facilities of the network data model. The network data model contains two basic data structures: records and sets. A record is a collection of named data items and a set is a relationship among records. As the basic unit of data manipulation, records may be stored, erased, found, modified, and connected or disconnected from other records. Sets maintain inter-record re- lationships and provide logical access paths so that a user may navigate from one record to the next.

The standard provides language facilities for defining records and sets and for defining applica- tion specific views of the data. In its current form, it does not provide syntax for declaring access control restrictions; instead, such restrictions are declared in a vendor specific syntax.

The N D L standard provides a Module Lan- guage for interface to other languages. Each NDL statement may be packaged as a procedure that can be called and have parameters passed to it from an external language. As each record is found, its data item values can be passed to the accessing language as parameters.

The schema definition language may be used to interchange database definitions and application specific views. The data manipulation language provides the data operations that make it possible to interchange complete application programs.

Applicability A database language standard is appropriate

for all database applications where data will be shared with other applications, where the life of the application is longer than the life of current equipment, or where the application is to be un-

Database management standards 189

derstood and maintained by programmers other than the original ones. The network data model, and thereby the NDL standard, is appropriate for highly structured, static database applications re- quiring rapid access along predefined paths. The network data model is less appropriate for flexible or ad hoc data retrieval and for this reason is gradually being replaced by relational and object- oriented database management systems.

Level of consensus The ANSI, ISO, and FIPS specifications are

essentially identical. The FIPS specification re- quires ANSI Level 2 conformance to one or more FIPS programming languages and requires a FIPS Flagger to flag extensions.

Product availability No known complete implementations of the

standard are commercially available. Instead, a number of vendors offer 'Codasyl compliant' sys- tems that incorporate some of the features of the standard and some of the functionality of an earlier Codasyl DDLC and Codasyl Cobol Jour- nal of Development. Syntactic differences in exist- ing products inhibit portability.

Completeness The existing standard specifies data definition,

data manipulation, and very limited integrity con- straints. The standard is complete for straight-for- ward applications, but lacks many desirable en- hancements, especially ad hoc query capabilities.

Maturity Network database implementations have ex-

isted for over twenty-five years. The technology is well-understood and many different proprietary products are based on network database topogra- phies. Evolution of existing products has essen- tially stopped as new database development ef- forts are going into relational and object-oriented systems.

Stability No changes to this specification are foreseen.

De facto usage The NDL standard will probably not be speci-

fied as a required part of a procurement, even if a network model database language is required. In-

stead, the standard will more often be used as a generic specification of desired features.

Limitations The maintenance overhead connected with ex-

isting implementations of network structured databases has caused many applications that use these systems to be redeveloped. The network implementations are often being replaced with relational database implementations conforming to Database Language SQL.

Conformance testing None available.

Future plans The existing ANSI and ISO NDL standard will

likely be withdrawn in 1992, or soon thereafter, unless some user or vendor establishes a require- ment for its continued existence.

Alternative specifications: Codasyl DDLC and Codasyl COBOL JOD.

4. Remote database access (RDA)

Existing specifications: None.

Emerging specifications: 1SO/IEC CD 9579- Partl Generic RDA (Expected late 1992)

ISO/ IEC CD 9579- Part2 SQL Special- ization (Expected late 1992)

Sponsoring organization:ISO/IEC JTC1/SC21

Description Remote Database Access (RDA) provides

standard protocols for establishing a remote con- nection between a database client and a database server. An emerging RDA standard will standar- dize distributed processing in a 'c l ient /server ' SQL environment, i.e. an environment with standard- conforming R D A / S Q L 'servers' at each remote node in a communications network. RDA speci- fies a two-way connection between a client and a server, as well as transfer syntax and semantics for SQL database operations. The client is acting on behalf of an application program or remote pro-

190 L. Gallagher

cess, while the server is interfacing to a process that controls data transfers to and from a data- base. The communications protocols are defined in terms of OSI standards for Association Control (ACSE), Remote Operations (RO), Transaction Processing (TP), and Commitment, Concurrency and Recovery (CCR). The goal is to promote distributed processing by standardizing the inter- connection among SQL database applications at nonhomogeneous sites.

The RDA specification is specified in two parts, a Generic RDA for arbitrary database connection and an SQL Specialization for connecting data- bases conforming to Database Language SQL. Both parts were registered as I S O / I E C Draft In- ternational Standard (DIS) in mid-1991 and are currently undergoing ISO/ IEC national body bal- lots for final adoption as international standards. Formal approval is expected in late 1992.

Generic RDA The proposed Generic RDA standard provides

an RDA Service Interface and an RDA Com- munications Element that exist at both the client and server sites. The Generic RDA Service Inter- face consists of service elements for association control, for transfer of database operations and parameters from client to server, for transfer of resulting data from server to client, and for trans- action management. Association control includes establishing an association between the client and server remote sites and managing connections to specific databases at the server site. Transaction management includes capabilities for both one- phase and two-phase commit protocols.

The RDA Communications Element at the cli- ent site converts RDA service requests into the appropriate underlying RO, ACSE, and TP proto- cols as part of an open systems interconnection. The RDA Communication element at the server site converts received protocols into requests to its underlying database management system.

The Generic RDA service does not specify the syntax or semantics of database operations sent from client to server. Instead, the standard as- sumes the existence of a language specialization (e.g. IRDS or SQL) that specifies the exact trans- fer syntax for standard operations.

SQL specialization The R D A / S Q L Specialization complements the

Generic RDA standard for use when a standard

conforming SQL data manager is present at the server location. The client site may also have an SQL conforming data manager, but this is not required. The client processor transforms user re- quests into the appropriate standard protocols for transmission across the network to the server site. SQL data operations are sent as character strings conforming to the SQL language and are packaged in an RDA/ASN.1 envelope that allows for em- bedded parameter values to be send with the data operation.

The result of an SQL statement will contain status code and exception code parameters and may contain one or more rows of data in response to a query. The transfer syntax for all such data is specified in ASN.1 as part of the SQL Specializa- tion standard.

Applicability Used to establish a remote connection between

a database client, acting on behalf of an applica- tion program, and a database server, interfacing to a process that controls data transfers to and from a database. The goal is to promote interconnection of database applications among heterogeneous eft- vironments, with emphasis on an SQL server in- terface, It is expected that the R D A / S Q L Special- ization will become the basis for all interconnec- tion among SQL database management products from different vendors. Interconnection among database products from the same vendor will likely continue to use vendor specific communication and interchange forms.

Level of consensus The I S O / I E C RDA specification is currently

undergoing DIS Ballot in JTC1/SC21. The speci- fication is in two parts: Part 1 - Generic RDA and Part 2 - SQL Specialization. Review copies should be available from OMNICOM as IS O / IEC DIS 9579-1 &2: 199x. RDA has been adopted as the remote database access component of SQL Access, X /Open , and other vendor consortia. RDA is also a working task group of the NIST OSI Implementor's Workshop.

Product availability There are no known existing RDA implementa-

tions, but many SQL vendors are planning to have conforming client and server products available before final adoption as a standard in the 1992

Database management standards 191

time frame. Vendor consortia, such as SQL Access and X/Open , hope to have working prototypes operational in 1991 to demonstrate interoperabil- ity among different SQL servers.

Completeness RDA services consist of association control,

data transfer, and transaction management be- tween a single client and a single server. Associa- tion control includes making a connection to a specific database at the server site. SQL state- ments are sent as character strings with a separate list of input parameters, and resulting data or exception conditions are returned. Transaction management includes capabilities for both one- phase and two-phase commit protocols. The exist- ing specification does not consider multiple simultaneous connections, so distributed database management is the concern of the client process. Extensions for true distributed database manage- ment among different SQL implementations are under consideration.

ing RDA is only a two-way service and protocol specification.

Conformance testing RDA will likely become a future optional part

of conformance testing for GOSIP. A the present time, RDA can be tested indirectly using the NIST SQL test suite, with application programs at the client site and data at the server site.

Future plans Enhancement projects for distributed database

and stored database procedures have already been proposed to ISO. Extensions to support new fea- tures in evolving SQL standards are under devel- opment.

Alternative specifications Vendor agreements reached by SQL Access,

X/Open , and other vendor consortia are finding their way into the RDA standard, but some agree- ments will always precede the formal standard.

Maturity Methods for establishing communications links

between client and server sites are well known, but agreements on non-proprietary communications protocols are very new.

Stability The cl ient/server architecture is just one of

several architectures used for implementing dis- tributed systems and there is no final conclusion as to which is best. The stability of RDA depends upon the stability of the cl ient /server architec- ture. At the present time the cl ient /server archi- tecture seems to be the architecture of choice for heterogeneous open systems interconnection.

5. Conclusions

Database management standards provide facili- ties for defining, managing, and protecting data. The SQL standard in particular is leading the way toward unprecedented portability and interoper- ability of database applications. The emerging RDA standard promises to complete the link among SQL products from different vendors, thereby leading to true open systems interconnec- tion. Emerging specifications for future revisions promise enhanced facilities to support object-ori- ented and knowledge-based applications in a dis- tributed processing environment.

De facto usage Since there are no existing implementations,

procurements have not yet used the RDA specifi- cation as a required component. Until delivery dates after 1992, proposals will likely be based on proprietary RDA implementations provided by individual vendors or consortia.

Limitations A common mistake is to assume that RDA is a

distributed database specification. Although dis- tributed extensions are under consideration, exist-

6. Organizations and standards groups referenced

ISO International Organization for Stan- dardization, recently entered into co- operative agreement with IEC via Joint Technical Committee 1 (JTC1) to jointly publish information tech- nology standards. All ISO publica- tions are available through ANSI. See OMNICOM for availability of early drafts.

| 92 L. Gallagher

IEC

A N S I

X3

N I S T

N T I S

In te rna t iona l Elec t ro technica l Com- mission, recent ly entered into cooper - ative agreement with ISO via Joint Technica l Commi t t ee 1 (JTC1) to

j o in t ly publ i sh i n fo rma t ion technol- ogy s tandards . A m e r i c a n N a t i o n a l S t anda rds In- sti tute, 1430 Broadway, New York, N Y 10018, publ ishes all A m e r i c a n N a t i o n a l S tandards , Sales phone : 212-642-4900, In te rna t iona l : 212-642- 4986. A n A N S I accred i ted commi t t ee for the d e v e l o p m e n t of I n f o r m a t i o n Processing S t anda rds in the Un i t ed States. X3 s t anda rds are pub l i shed through A N S I . See G L O B A L for avai labi l i ty of ear ly draf ts . Na t iona l Ins t i tu te of S t anda rds and Technology, develops s t anda rds and guidel ines (FIPS) for the U n i t e d States federal government for sale through NTIS. N a t i o n a l Technica l I n f o r m a t i o n Service, Springfield, VA 22161, pub- lishes all F I P S document s , Sales phone: 703-487-4650.

G L O B A L G l o b a l Engineer ing, Inc. A com- merc ia l f i rm that has a special rela- t ionship wi th X3 to provide copies of draf t p r o p o s e d A N S I / X 3 s t anda rds before formal publ ica t ion , phone: 800-854-7179.

O M N I C O M O M N I C O M , Inc. A commerc ia l f i rm that has a special re la t ionship wi th I S O / I E C JTC1 to p rov ide copies o f d r a f t p r o p o s e d I n t e r n a t i o n a l S t anda rds before formal publ ica t ion , phone: 800-666-4266.

Leonard Gallagher is a computer sci- entist in the Information Systems En- gineering Division at NIST. He is re- sponsible for data models, database standardization, and integration of database technology with new ap- proaches to information management such as knowledge and object-oriented systems and hypertext. He has been a member of the ANSI/X3 technical committee on Database, X3H2, since 1979 and chairs the international de- velopment group for follow-on en-

hancements to the ISO SQL standard. At NIST, he also leads an inter-division project for research and experimentation into hypertext and hypermedia information management systems. Dr. Gallagher received the BA degree in mathematics from St. John's University of Minnesota in 1965 and the Ph.D. in mathematics from the University of Colorado in 1972. He taught mathematics at the Catholic University of America for 6 years and has been involved in database research at NIST for the past 13 years.