Background: Web-Servers and CGI - Superbase Internet Webserver

Before building a web site, it is important to understand the architecture of the Internet. The World Wide Web (WWW) is composed of three principle components: A wide-area network running TCP/IP; web-servers; and web-browsers. TCP/IP is a platform independent (Windows, Mac, OS2, UNIX, etc.,) network protocol used by the Internet (and Intranets). It defines a standard for sending and routing packets of data between servers and clients. In this case, web-servers and web-browsers. A web-server is a simple application that waits for requests from a web-browser and returns files (HTML, images, etc.). A web-browser is also a simple application that displays information (decodes the HTML and inserts/formats the appropriate images) in a window.

HTML files are simple ASCII files which are used to hold information that will appear within browsers. These text files contain embedded codes for font attributes, locations of images, simple form elements (edit boxes, list-boxes, check-boxes, buttons, and radio buttons), etc.

There are only a few dozen formatting elements available for HTML 3.2, although several advanced extensions such as frames, tables, and colors, are supported by proprietary browsers from Microsoft and Netscape.

The architecture of this system relies on page-oriented hyper-text links to structure and disseminate information. Graphics are being increasingly used by web page designers to provide the illusion of a more cohesive document structure. Although at its roots the system is considerably less sophisticated than most people realize. Perhaps the most important consideration when designing web content is the static nature of the medium. There is no way to interactively communicate with a server to update the contents of a web page without resorting to Java on the client/browser side.

The standard method of requesting information from a web-server is to use a URL (Uniform Resource Locator). This is usually the address of an HTML file on disk. i.e. "http://www.geometrix.bc.ca/geonew1.htm". If this URL were entered in a web-browser's "Open Location" edit box, it would cause the GeoMetrix web-server to return the "What's New" page from the GeoMetrix web site.

For web-servers that support CGI (Common Gateway Interface), nearly all do, you may also send a URL that contains the name of an executable (.exe, .com, or .bat) file. The web-server will run the application, passing in names of files containing posted form information, and wait for it to terminate. The application is responsible for creating an HTML file which the server will return after the application terminates.

Executables can be as simple as a batch file. However accessing a database is a far more complex operation. Even with C/C++, Pascal, or Delphi, the overhead of loading a 500K+ executable (and/or DLLs) and connecting to a database can take far too long for the server to process each transaction. Larger database applications, such as Superbase, FoxPro, Paradox, dBase for Windows, Access, etc. could not be repeatedly loaded in such a manor. Such applications normally use a small intermediate executable (Gateway Program) which, when run, contacts a copy of FoxPro, Paradox, etc., that is already running. This can be accomplished through a number of methods the most common being DDE.

For example, connecting to a copy of Superbase using a web-server and CGI can be accomplished as described in Figure 3.

Figure 3: CGI Interface to Superbase

This configuration processes each transaction request sequentially. Transactions are usually completed within a fraction of a second allowing up to several hundred to be completed each minute (on Pentium-based servers). Our experience has shown that when using this configuration the only bottleneck that occurs is with the web server software trying to send back the output to slow connections (a client browser using a 14.4 connection through a slow ISP). This in no way limits the ability for Superbase to continue processing transactions.