WWW Servers

WWW servers provide information to the Web. Server software is available for many computer platforms, but setting up a server isn't always easy.

Why Set Up a WWW Server?: Even if you don't have an Internet connection, there are lots of uses for an internal WWW server.
WWW Server Design: Setting up a server to provide information to the many different Internet clients requires extra thought, but the effort is worth it.
Setting Up Your WWW Server: Server software exits for UNIX, MPE, Windows NT, Microsoft Windows, and even MS-DOS.
Maintaining Your WWW Server: Like most applications, your WWW server will need a little help from time to time.

Why Set Up a WWW Server?

If you have a full-time Internet connection, you might want to set up a WWW server to provide information about your company, your division, your group, or yourself. Even if you are not connected to the Internet, you still might want to set up a server.

Hypertext is a useful way to distribute information because it can contain mixed text and graphics (or more), as well as links to other documents. Using WWW servers, you can create sophisticated help systems without a lot of work. Once established, these systems then become available to all users on your internal network who have suitable client software (browsers).

With CGI scripts and e-mail, you can automate forms which you now process by hand (e.g., expense reports, travel reports, or purchase requisitions). With some extra work, you could even have the forms processed directly into a database. You can also design scripts to look up information in your existing databases and display it for clients.

If your users are pushing for Microsoft Windows interfaces to all of their database data, you can use your WWW server as an intermediate solution. This way users get an immediate graphical interface and managers can experience the difficulties of managing client/server configurations.

WWW Server Design

When you set up a WWW server, keep in mind that many different clients will be accessing your server. If your server is available on the Internet, you should not assume that the clients will all have high-speed Internet connections and graphical browsers.

Consider these things when designing your WWW server:

Concentrate on your text. Well-written text conveys a lot of information. If you use text to convey essential information, then your server will be friendly to text-based clients like Lynx.
Organize documents the way you would organize a book: gather information together into chapters; each chapter should describe a single idea or related topics. Provide navigational tools (like previous or next chapter) and an overview with a table-of-contents. We have attempted to have all of these elements in this paper.
Question each graphical image that you provide. Does the graphic add meaning to the text or is it just neat? Compare the size of the graphics file to the size of your text files. If the graphical image is much larger, does it really add a lot of necessary information?
If your WWW server is on a fast network, do all the clients have fast access to your server? You may have a T1 connection (1.544M bits/sec), but many WWW clients connect via 14.4 modems. Some commercial Internet providers even charge by the hour, which makes it more expensive for clients to download large files and graphics. If your clients have a fast connection to the Internet, you can provide more graphical information and larger text files without annoying them. Nevertheless, it's a good idea to keep these limitations in mind when you're developing your server.
Try to keep files to a reasonable size (we suggest three to ten thousand bytes long). When converting existing documents to HTML, remember that they will often end up quite large (tens of thousands of bytes). Do clients want to download such a large file only to find that it is of no interest? The converse is also true. Can clients download a single file with the complete text (e.g., this paper), without having to follow all the hypertext links?
Hypertext does not mean disorganization. Provide an index or a table of contents to your web pages, so users can quickly find information. Provide summaries for long articles and files.
Use graphic-design common sense. Use white space to increase readability. If you use special effects (bold, italics, underline, horizontal rules, etc), use them sparingly to increase their effect.
If your WWW server is available on the Internet, many visitors will access your server out of curiosity. Make your welcome page attractive, but clearly identify what information your WWW server is providing. Of all the files you publish, be most careful of the size of your welcome page. It will likely be the most frequently accessed page.

We also suggest that you look at the W3 Style Guide.

Setting Up A WWW Server

First, you need to decide what computer will host your WWW information (or you could pick several hosts). If your WWW server will make information available to many machines, the host must be connected to your network or the Internet.

While WWW server software is available for a variety of machines, each server software package runs only on certain operating systems. The server software you pick will have to be compatible with the host machine that provides the WWW service.

WWW Server Software

W3 maintain a good list of WWW server software. Two of the most popular UNIX WWW server software packages are NCSA HTTPD and CERN HTTPD. A pre-compiled copy of the NCSA HTTPD software is available for MPE/iX.

Windows NT is becoming more popular as a WWW server, largely due to its built-in networking support and its familiar Windows interface. Free Windows NT HTTP Server software is available from the European Microsoft Windows NT Academic Center. The Robelle Windows NT WWW Server uses the O'Reilly Website software. Website comes with comprehensive documentation -- something other server software is lacking.

Configuration and management is different for each package. We found the O'Reilly Book Managing Internet Information Services to be a valuable resource in setting up our WWW servers. The book is an excellent introduction to HTML, with many good examples of configurations. Unfortunately, the book only covers the configuration of the NCSA HTTPD software.

Security

The CERN and NCSA HTTPD packages allow the WWW administrator to configure security. By default, both packages allow anyone to connect to your WWW service. However, you can configure the servers to allow connections only from specific IP addresses (be sure to do this if your WWW service is for internal use only). You can also password protect individual files. The MPE WWW Server includes a demonstration of the NCSA security features.

By default, the CERN and NCSA server software allow individual directories of hypertext files. If someone specifies a URL with a directory starting with tilde (~), the server software looks for a user directory of that name and then searches under the user name for the directory public_html.

Writing HTML

Once you have the WWW server software running, you need to create WWW information. WWW documents use the Hypertext Markup Language (HTML). See the HTML description earlier in this paper for suggestions and tools for writing HTML.

Be sure to test your files before adding them to your WWW server. We test with at least three different browsers (Lynx, Mosaic, and Netscape). We also use Weblint on all of our Web documents. Weblint checks for common errors in HTML. While Weblint isn't perfect, it does help produce HTML that is acceptable to the widest range of WWW browsers.

Weblint is written in Perl. To use Weblint, you must have a working copy of Perl. Perl is short for "Practical Extraction and Report Language". Perl is designed to be more powerful than the shell, but easier to use than C.

Host Name

If your WWW server is available on the Internet, it's a good idea to create an alias for the actual computer that hosts your WWW service. Most people chose "www" as the alias name. This will make it easier for you to change the host without affecting users of your WWW service.

Robots

WWW servers on the Internet are often visited by robots. Robots usually visit Web sites in order to create indexes of the information that you publish on your WWW server. Since robots can cause problems for a WWW server, it's a good idea to create a robots.txt file. This file tells well-behaving robots which parts of your WWW they should visit. You might want to exclude graphical images, CGI scripts, and forms from a robot search, but include all other information about your WWW server.

Internal WWW Servers

If your WWW server will only be available on a Local Area Network, you have more flexibility in your design. Since users will have reasonably fast access to the server, you can make your HTML pages larger. You can also distribute more binary objects, such as graphics, word-processing documents, and spread sheets. You do have to remember to configure each client browser with the information on how to handle each filename suffix (e.g., you might want to associate ".doc" with Microsoft Word). See the section on External Viewers in the Clients section of this paper for more information.

Maintaining Your WWW Server

Once you have your WWW server working, you need to continue maintaining it. The Web is changing rapidly. You need to insure that you obtain newer versions of the HTTPD software from the original source.

All WWW server software can produce log files. If you do enable log files (some software has them enabled by default and others not), they usually grow without bounds. At Robelle, we make a copy of the current log files once a day and then we empty them. We keep the daily copies for approximately 60 days. This lets us provide statistics about our WWW service through the getstats program.

Because more and more users are joining the Internet, you will likely want to continue to improve and expand your WWW information. This is a challenge, since the conversion and authoring tools are not yet well developed. At Robelle, we have tried to automate some of the production of our WWW information. For example, when the most recent change notices for Qedit/MPE, Qedit/UX, Suprtool/MPE and Suprtool/UX are released, they are automatically posted to the Robelle FTP Service