WWW5 tutorial: collecting and serving information

introduction:

this tutorial covered exactly the problems we are currently facing at the ETH in general and the "informatikdienste" in particular: the pros and cons of having multiple independent WWW servers versus one centralized WWW server.
of course, there is no way to centralize ALL HTML documents currently available at the ETH in ONE PLACE and it would not make any sense at all to try to do so. nevertheless, there are now discussions about subjects addressed by this speech such as:

• uniform style over all documents
• ease of navigation within the organization's information
• searching and indexing
• consistent and persistent URLs
• valid HTML syntax

different models of WWW server structures:

there are various ways to structure one or more WWW servers within one organization. the two major approaches are "distributed structures" and "centralized structures". in this tutorial, they distinguished the following five structures ranging from "anarchy" to "dictatorship" with the first two structures considered "distributed":

1. multiple independent servers
2. multiple coordinated servers
3. generally accessible central server
4. subscription server
5. controlled centralized server

three of the above models (1, 3 and 4) have been discussed in greater detail and the advantages and disadvantages were compared from the user's, the information provider's and a global point of view. at the ETH, we basically have dozens of more or less independent servers (there is some minimal coordination particularely in the addressing scheme of the servers), but at the same time, we run some generally accessible central servers ( APACHE, ezInfo, WAWONA) that are used by many different organizational units which do not run their own WWW server for various reasons.

multiple independent servers:

many organizational units inside the organization run their own WWW server. at the ETH, this organizational units are departments, institutes, working groups etc. these organizational units operate their server on their own, but can get support from the "informatikdienste" if they want to.

advantages:

• the information providers can choose their most favorite platform (hardware, operating system, httpd daemon, etc.) and operate the server the way they like it best.
• users benefit from multiple servers because they do not represent a single point of failure and they may like shorter URLs better.
• in general, the organizational units do not depend on a central service.

disadvantages:

• because the information providers run their own server, they have to act as WebMasters (taking care of the server software, etc).
• information may be duplicated due to the lack of coordination.
• the structure of the URLs may be inconsistent within the organization.
• URLs may be more non-persistent as changes may occur more frequently.
• searching over all Web servers of the organization is much more difficult.
• it may be more difficult to apply an uniform style through the organization's Web pages.

generally accessible central server:

the information providers use a central server to create and maintain their own documents. at the ETH, the "informatikdienste" operate three central servers that are used by various organizational units to provide their Web pages. once the so called Web moderator has received a username and a password to access the server, she or he create and maintain her/his documents on her/his own responsibility. most providers create the documents on their own workstation or PC and transfer the documents to the server when they are ready to be published. others may create and maintain the documents directly on the server. it is also possible to use NFS or AFS so that the provider can maintain the documents on her/his private workstation while the files still reside on the server. another convenient way would be the use of an authoring tool that supports remote save.

advantages:

• the provider can benefit from a central server because she or he does not have to deal with the server software itself. also data backup and all other management tasks will be covered by a central service. in addition, the central server may provide other services, such as cgi scripts, HTML syntax and link checking, generation of statistical information and so on.
• the user may benefit from a centralized search index and more consistent presentation of the information.
• in general, system management efforts may be reduced if only one or at least a small number of servers have to be maintained.

disadvantages:

• information providers must rely on a central service and may have to follow some policies.
• a single server represents a single point of failure. if the server is down, no information about the whole organization will be available to the Web.

subscription server:

a subscription server is a centralized server which does not give information providers direct access to their Web pages. instead, it provides means to transfer data to the server, usually via HTML forms. documents may not only be transferred from the provider's system to the central server, but they may also be checked, converted and registered in a database. currently, we do not operate a subscription server at the ETH.

advantages:

• information providers may submit their documents in their most favorite format, for example as postscript file, word document or even as a printed paper. the team operating the central server will convert the documents to the appropriate format, e.g. HTML, PDF etc.
• because all documents are processed somehow, they may all have a unified layout, the links will be more consistent and persistent and from the registration database, users may get valuable information about the documents and the authors.
• searching and indexing is much easier.

disadvantages:

• information providers have no direct access to their Web pages and updates may take longer because they have to be processed.
• if documents are stored in different formats, they need a lot of disk space. if conversion is done on the fly, access to documents may be slow.

conclusion:

all of the above mentioned structures have their advantages and disadvantages. in an organization such as the ETH, a mixture of distributed and centralized servers is probably the only practical solution.
in my opinion, centralized servers are good for:

• organizational units that cannot or do not want to operate their own server, either because they lack people with the required knowledge, because they do not want to spend time to manage a server or because they prefer to rely an a system that is managed by professionals.
• important documents that shall be presented in a consistent style, such as the homepage of the organization itself and homepages of important and highly visible organizational units, such as public relation.
• documents that do not change very often and which URLs shall be short, simple and persistent.

on the other hand, pages that changes very often shall be located on the organizational unit's server. this is particularly true if the information is dynamically created either from a database or by a cgi script.

information shall never be duplicated, it should always be available from its source !

the slides of this presentation are available on the Web.

back to WWW5 main document.