Database mix poses a sharing challenge; IS execs build ties when none available; find retrieval easier than distributed update - Client/Server Computing - includes a related article on important client/server issues

Software Magazine, March, 1992 by Barbara Bochenski

Ansi-compliant SQL frees the programmer from having to worry about the differences in SQL dialects. The widespread use of TCP/IP and common RPC mechanisms in the Unix world means programmers do not have to be too concerned with how a request is going to get the server, he said.

Despite the advantages of a Unix environment, Vaughn maintained that in the public's mind, the term "client/server" is more associated with non-Unix systems such as OS/2. "Furthermore, the OS/2-based database servers have been more widely publicized and are considerably less expensive," he explained.

A typical environment has a mix of operating systems installed on desktop, server and host machines. From a programming perspective, the mix of different operating systems, database systems and network operating systems from separate vendors present complexities.

"Each competing network operating sysgem--Microsoft's LAN Manager, IBM's LAN Server and Novell's NetWare--have completely different transport mechanisms and completely different transaction protocols," said Vaughn. "For example, Novell is IPX/SPX-based, LAN Manager uses Named Pipes, and LAN Server is a mixture of Named Pipes and APPC."

He noted that the same thing applies to database vendors. "One database responds to SPX transactions, another responds to Named Pipes calls, while others respond to NetBios or APPC calls. If a program were written to directly communiate with each database," said Vaughn, "it would have to use three different APIs [application program interfaces] just to establish a link with those servers over the LAN."

George Zagelow, manager of architecture and standards at IBM's Santa Teresa Labs in San Jose, Calif., said a major thrust of distributed database work is to make it easier to access data from remote databases. He cited three primary architectures available for sharing data: Distributed Relational Database Architecture (DRDA), Remote Data Access (RDA) and the SQL Access Group (SAG) approach.

"DRDA is ours," said Zagelow. "We have products rolling out for DRDA in early 1992. The whole idea behind DRDA, RDA and SQL Access is to hide the communications behind SQL--to make it easy to get data from remote databases."

There is a different, however, in the way each of the three architectures accomplish this. For sites dealing with separate products that each use a different one of these architectures, accessing data may not be so easy after all.

Last fall Microsoft announced Open Database Connectivity, which is based on a specification from the SQL Access Group, an industry consortium consisting of most of the major database vendors except IBM.

SQL has been implemented in more diverse ways in the DOS-Windows-OS/2 world than in the Unix world, Vaughn said. "This requires that the SQL statements for each database be incorporated in the program in different way." Vaughn summarized two different approaches: one used by IBM and another used by Sybase.

"IBM's Data Manager uses embeded SQL," he explained. "The actual SQL code is incorporated in the Cobol code." The program must be processed with an IBM precompiler which is used to compile the SQL code separately.