The history of the SQL language is intimately intertwined with the development of relational databases. Table 3-1 shows some of the milestones in its 30-year history. The relational database concept was originally developed by Dr. E.F. “Ted” Codd, an IBM researcher. In June 1970, Dr. Codd published an article entitled “A Relational Model of Data for Large Shared Data Banks” that outlined a mathematical theory of how data could be stored and manipulated using a tabular structure. Relational databases and SQL trace their origins to this article, which appeared in the Communications of the Association for Computing Machinery.
1. The Early Years
Codd’s article triggered a flurry of relational database research, including a major research project within IBM. The goal of the project, called System/R, was to prove the workability of the relational concept and to provide some experience in actually implementing a relational DBMS. Work on System/R began in the mid-1970s at IBM’s Santa Teresa laboratories in San Jose, California.
In 1974 and 1975, the first phase of the System/R project produced a minimal prototype of a relational DBMS. In addition to the DBMS itself, the System/R project included work on database query languages. One of these languages was called SEQUEL, an acronym for Structured English Query Language. In 1976 and 1977, the System/R research prototype was rewritten from scratch. The new implementation supported multitable queries and allowed several users to share access to the data.
The System/R implementation was distributed to a number of IBM customer sites for evaluation in 1978 and 1979. These early customer sites provided some actual user experience with System/R and its database language, which, for legal reasons, had been renamed SQL, or Structured Query Language. Despite the name change, the SEQUEL pronunciation remained and continues to this day. In 1979 the System/R research project came to an end, with IBM concluding that relational databases were not only feasible, but could be the basis for a useful commercial product.
2. Early Relational Products
The System/R project and its SQL database language were well-chronicled in technical journals during the 1970s. Seminars on database technology featured debates on the merits of the new and heretical relational model. By 1976, it was apparent that IBM was becoming enthusiastic about relational database technology and that it was making a major commitment to the SQL language.
The publicity about System/R attracted the attention of a group of engineers in Menlo Park, California, who decided that IBM’s research foreshadowed a commercial market for relational databases. In 1977 they formed a company, Relational Software, Inc., to build a relational DBMS based on SQL. The product, named Oracle, shipped in 1979 and became the first commercially available relational DBMS. Oracle beat IBM’s first product to market by a full two years and ran on Digital’s VAX minicomputers, which were less expensive than IBM mainframes. The company aggressively sold the merits of the new relational style of database management, and eventually renamed itself after its flagship product. Today, Oracle Corporation is the leading vendor of relational database management systems, and a major vendor of enterprise applications based on the Oracle database, with annual sales of many billions of dollars.
Professors at the University of California’s Berkeley computer laboratories were also researching relational databases in the mid-1970s. Like the IBM research team, they built a prototype of a relational DBMS and called their system Ingres. The Ingres project included a query language named QUEL that, although more structured than SQL, was less English-like. Many of today’s database experts trace their involvement with relational databases back to the Berkeley Ingres project, including the founders of Sybase and Illustra (now owned by IBM), and many of the object-oriented database startup companies.
In 1980, several professors left Berkeley and founded Relational Technology, Inc., to build a commercial version of Ingres, which was announced in 1981. Ingres and Oracle quickly became arch-rivals, but their rivalry helped to call attention to relational database technology in this early stage. Despite its technical superiority in many areas, Ingres became a clear second-place player in the market, competing against the SQL-based capabilities (and the aggressive marketing and sales strategies) of Oracle. The original QUEL query language was effectively replaced by SQL in 1986, a testimony to the market power of the SQL standard. By the mid-1990s, the Ingres technology had been sold to Computer Associates, a leading mainframe software vendor.
3. IBM Products
While Oracle and Ingres raced to become commercial products, IBM’s System/R project had also turned into an effort to build a commercial product, named SQL/Data System (SQL/DS). IBM announced SQL/DS in 1981 and began shipping the product in 1982. In 1983, IBM announced a version of SQL/DS for VM/CMS, an operating system that was frequently used on IBM mainframes in corporate information center applications.
In 1983, IBM also introduced Database 2 (DB2), another relational DBMS for its mainframe systems. DB2 operated under IBM’s MVS operating system, the workhorse operating system used in large mainframe data centers. The first release of DB2 began shipping in 1985, and IBM officials hailed it as a strategic piece of IBM software technology. DB2 has since become IBM’s flagship relational DBMS, and with IBM’s weight behind it, DB2’s SQL language became the de facto standard database language. DB2 technology has now migrated across all IBM product lines, from personal computers to network servers to mainframes. In 1997, IBM took the DB2 cross-platform strategy even farther, by announcing DB2 versions for computer systems made by Sun Microsystems, Hewlett- Packard, and other IBM hardware competitors.
IBM made another major stride in its cross-platform strategy in 2001, when it acquired Informix’s database business, and especially Informix’s installed base on non-IBM UNIX-based servers. According to most industry analysts, IBM is the second-largest vendor of database management software, and some user surveys actually placed it first, slightly ahead of Oracle in market share.
4. Commercial Acceptance
During the first half of the 1980s, the relational database vendors struggled for commercial acceptance of their products. The relational products had several disadvantages when compared to the traditional database architectures. The performance of relational databases was seriously inferior to that of traditional databases. Except for the IBM products, the
relational databases came from small upstart vendors. And, except for the IBM products, the relational databases tended to run on minicomputers rather than on IBM mainframes.
The relational products did have one major advantage, however. Their relational query languages (SQL, QUEL, and others) allowed users to pose ad hoc queries to the database—and get immediate answers—without writing programs. As a result, relational databases began slowly turning up in information center applications as decision- support tools. By May 1985, Oracle proudly claimed to have over 1000 installations. Ingres was installed in a comparable number of sites. DB2 and SQL/DS were also being slowly accepted and counted their combined installations at slightly over 1000 sites.
During the last half of the 1980s, SQL and relational databases were rapidly accepted as the database technology of the future. The performance of the relational database products improved dramatically. Ingres and Oracle, in particular, leapfrogged with each new version claiming superiority over the competitor and two or three times the performance of the previous release. Improvements in the processing power of the underlying computer hardware also helped to boost performance.
Market forces also boosted the popularity of SQL in the late 1980s. IBM stepped up its evangelism of SQL, positioning DB2 as the data management solution for the 1990s. Publication of the ANSI/ISO standard for SQL in 1986 gave SQL official status as a standard. SQL also emerged as a standard on UNIX-based computer systems, whose popularity accelerated in the 1980s. As personal computers became more powerful and were linked in local area networks (LANs), they needed more sophisticated database management. PC database vendors embraced SQL as the solution to these needs, and minicomputer database vendors moved down market to compete in the emerging PC local area network market.
Through the early 1990s, steadily improving SQL implementations and dramatic improvements in processor speeds made SQL a practical solution for transaction processing applications. Finally, SQL became a key part of the client/server architecture that used PCs, local area networks, and network servers to build much lower-cost information processing systems.
SQL’s supremacy in the database world has not gone unchallenged. By the early 1990s, object-oriented programming had emerged as the method of choice for applications development, especially for personal computers and their graphical user interfaces.
The object model, with its objects, classes, methods, and inheritance, did not provide an ideal fit with the relational model of tables, rows, and columns of data. A new generation of venture capital-backed “object database” companies sprang up, hoping to make relational databases and their vendors obsolete, just as SQL had done to the earlier, nonrelational vendors. However, SQL and the relational model more than withstood the challenge. Total annual revenues for object-oriented databases are measured in the hundreds of millions of dollars, at best, while SQL and relational database systems, tools, and services produce tens of billions of dollars of sales per year.
As SQL grew to address an ever-wider variety of data management tasks, the one- size-fits-all approach showed serious strain. By the late 1990s, database management was no longer a monolithic market. Specialized database systems sprang up to support different market needs. One of the fastest-growing segments was data warehousing, where databases were used to search through huge amounts of data to discover underlying trends and patterns. A second major trend was the incorporation of new data types (such as multimedia data) and object-oriented principles into SQL. A third important segment was mobile databases for portable personal computers that could operate when sometimes connected to, and sometimes disconnected from, a centralized database system. Another important application segment was embedded databases for use within intelligent devices, such as network equipment. In-memory databases emerged as another segment, designed for very high levels of performance.
Despite the emergence of subsegments of the database market, SQL has remained a common denominator across them all. As the computer industry prepares for the next century, SQL’s dominance as the database standard remains very strong. New challenges continue to emerge—databases rooted in the eXtended Markup Language (XML) are the latest attempt to move outside of the relational model and SQL—but the history of the past 20 years indicates that SQL and the relational model have a powerful ability to embrace and adapt to new data management needs.
Source: Liang Y. Daniel (2013), Introduction to programming with SQL, Pearson; 3rd edition.