The SQL Standard: How it Happened

When I left a faculty position at the University of Maryland to work full time for Dennis Fife at the US government agency now called the National Institute of Standards and Technology (NIST),¹ we agreed that I could spend 50 percent of my time on my doctoral dissertation research on modeling and measurement of database management system (DBMS) performance. As part of the other 50 percent, Dennis directed me to join 20 others from 17 different organizations at the first meeting of the X3H2 Database Standards Committee (DSC)² in Washington DC in April 1978. Initially chartered only to develop a standard Data Definition Language (DDL) for network DBMSs, in time the DSC became the epicenter of US and international SQL standardization. I became the DSC vice chair at the second meeting and chair at the committee's 11th meeting in May 1980 when the first chair John Rundell resigned. Throughout my career that included (in addition to NIST) GE, Sybase, and now Oracle, I have continued to chair the DSC. It is from this perspective that I describe my view of how the SQL standard happened.

1978: A Different IT Marketplace

The IT market in the late 1970s was substantially different than it is today. DBMS and other system software were generally bundled with and acquired along with hardware from IBM or one of the hardware system companies. IBM's market-leading product was IMS, a hierarchical product with network features at the leaf nodes. Some other system vendors provided DBMS products based on specifications developed by Codasyl, the group that created Cobol, the most widely used commercial programming language. Codasyl DBMSs structure data in owner-coupled sets using embedded pointers. In addition to the hardware system companies, a growing number of independent software vendors were providing implementations of the Codasyl specifications (IDMS from Cullinet), hierarchical systems (System 2000 from MRI), and various types of other DBMS products using shared-value relationships (ADABAS from Software AG).

All the hardware system companies and many independent software vendors participated in the early DSC efforts. Because the US government was then an active player in IT standardization, as the NIST representative, I was cast as the dominant buyer in opposition to IBM, the dominant provider. Undoubtedly, some participants were at the table primarily to get early insight into how the NIST/IBM tug of war resolved.

Codasyl DBMS

Consistent with its limited initial charter, DSC completed a DDL specification for Codasyl network DBMSs in 1982. When the Data Manipulation Language (DML) specifications needed to exercise the DDL were not available from Cobol, Fortran, or any other ANSI-accredited programming language committee, however, there was no reason to submit the DDL working draft for approval as an American national standard (ANS). Instead, DSC sought and obtained authority to produce a full-function DBMS specification—one including DML as well as DDL.

Development of a "complete" Codasyl DBMS standard, which was named Database Language NDL (Network Database Language) moved slowly as the committee worked to make it better rather than just good enough. The largest contributor to NDL was IBM's Phil Shaw, who not only edited the working draft, but also contributed many of the technical papers that substantially improved the specification. Of course, the vendors of non-Codasyl DBMS products were not unhappy that development of the NDL standard was taking longer than expected.

Relational Database Model

Stimulated by the seminal papers by E.F. (Ted) Codd in the early 1970s, university and industry research was focused on a new mathematically based approach to DBMS that did not require predefinition of relationships. Work on this "relational" approach moved from academic papers to development of prototypes, including IBM's System R using SQL and the University of California, Berkeley's INGRES, which used the QUEL language. Even as they were contributing to the development of a Codasyl-based standard, a few of the vendors participating in the DSC were developing relational database management system (RDBMS) products. And a young entrepreneur named Larry Ellison was deciding to go out on his own to build what became the first commercial implementation of SQL...