Designing and Building the CS Departmental Server.

By Jack Hsu


When a person, say in California, wants to know more about our department, the choices that this person have are to call up our department, to send in a postcard requesting for a brochure, or to fly in for a campus visit. For the first option, the telephone conversation will probably last for not more than half an hour and the person will not get a complete picture of our department, not to mention the fact that some ideas are hard to convey through words. The second option will require the person to wait for at least two weeks for something that may or may not answer the questions that the person may have. The last option will require the person to invest a substantial amount of time and money.

It is with these issues in mind that Professor Bill Pugh initiated a project to build a departmental information server accessible from the internet that will contain general information of the department, as well as on-line technical papers.


The departmental server should satisfy the following requirements:
  1. Easily accessible. The user should not be required to have exotic equipment or software in order to access the server.
  2. Flexible. It should be able represent at least text and graphics, and allow the user to initiate searches on certain indexes, such as research paper topics.
  3. User Friendly. The interface should be easy to use and intuitive.
  4. Complete. It should contain enough information to answer any question a person may have on the department.
  5. Well organized. It should be clear to the user where to look for a specific information.


The process of designing and implementing the departmental server can be divided into three subparts.

a. Choosing the Information System.

Deciding which information system to use was easy. Gopher and WAIS, the two most widespread systems used today, are both text-based systems. Moreover, Gopher does not have text formatting capabilities. WAIS, on the other hand, does not have links to follow from one document to another related document.

The World Wide Web (WWW) provides all the features of Gopher and WAIS and more. The WWW is a distributed hypermedia system. This means that WWW browsers can display text files as well as images, sounds and animations. This fulfills requirement 2 from above.

Although the WWW software was not as deeply deployed as WAIS and Gopher software, it is gaining in popularity, with the number of users doubling every 4 months. In fact, in June 1993 WWW's own http protocol moves ahead of the WAIS in the NSF Backbone packet count statistics. (Click here to look at the statistics.) The WWW software is becoming a standard package on filesystems. Our very own computer science department provides WWW clients on its filesystems.

In addition, there are WWW clients on a wide variety of platforms, including the Macintosh, Microsoft Windows, and X capable UNIX workstations, all commonly available hardware in a university and other environments. A user with a text-only terminal can also access a WWW server, although only the text will be visible. This satisfies requirement 1.

WWW clients such as the popular NCSA xmosaic has a graphical interface that is simple and easy to use, and the only interaction required by the user is to 'point and click and to type in search keywords. No training is required to use this client. This satisfies requirement 3.

b. Designing the Structure.

The organization of the information was designed with one intention in mind: to be clear and concise. Users should be able to find what they are looking for at one glance. We studied the organization of materials from other schools ( Stanford, Yale, CMU, and Indiana) and improved on what we thought was lacking from their designs.

This is the organizational plan that we came up with:

Campus Education * Undergraduate brochure * Undergraduate course * Graduate brochure News & Events Faculty and Research projects * By group: * AI * Computer Networks * Databases * Distributed Systems * High Performance Computing * Human Computer Interaction and Graphics * Numerical Analysis * Programming Languages * Software Engineering * Theoretical Computer Science * Listing of all professors Research papers * Listing of all papers * Keyword search of Maryland papers * Global search (across many CS departments) Ftp Student ACM Page DC Area Computer Science Research pointers Net Resources The faculty pages and the research paper pages are interlinked, so you can go to any faculty page from a research paper page that the faculty authored, or alternatively go to any research paper pages written by a faculty from that faculty's page. Most other schools offer only a linear one way search.

The main problem we feel that exists across servers of other schools is that their information appears to be organized in a haphazard way. There is no clear structure, which may be due to in part by an incomplete collection of information. We wish to avoid this in our server by defining a clear hierarchical structure with a broad but related range of topics.

Some schools (click here to look at Yale University) use a lot of graphics in their server. While this makes the pages more visually appealing, we feel that they do not improve the clarity in the organization of the materials. In fact, too many pictures embedded in the text actually make the system more confusing, especially when the pictures bear little or no relation to the text. We agree that pictures should be used only where appropriate in our server. That should include menu buttons, headers, and in the picture sections of our campus and DC area. It should not appear embedded in the text.

We also decided to use a list format rather than a paragraph format for the main menu (Click here to look at CMU) as our structure. Once again, this decision is made for the sake of clarity. While a paragraph format may have the advantage of being more descriptive, we think that we can circumvent this problem by providing descriptive texts in the pages linked from the main menu.

c. Implementation and Current Progress.

The server now runs by using the FTP protocol. After all the HTML pages are completed, a WWW server will be used.

The server as of now is unfinished, and the development will go on until early next semester. Also, you can look at the experimental home page by Professor Pugh.


It is my hope that our departmental WWW server will meet Professor Shneiderman's expectation of being the best there is. If you have any suggestions on how to improve the WWW server, please send me email (

Distribution of Work and Credits

Professor Pugh and Jack Hsu came up with the design.

Jack Hsu implemented everything and wrote this report.

Many thanks to Professor Ben Shneiderman for his guidance and helpful suggestions, and for providing the challenge and inspiration of "being the leader in the field". That shall always be the goal I strive for in my future endeavors.

Jack Hsu