The use of the Web browser as a platform independent, legacy system front end, is still in its infancy. When existing systems are required to be made more open, i.e., in terms of increasing access to newly empowered users, the use of Web browsers is one of the most attractive options.
This is because web based applications provide a common look and feel, regardless of the computing platform or geographical location.
The Internet is increasingly being seen, by the key players in the provision of the Information Superhighway, as the service delivery medium of the future [HREF 2] . Consequently web browsers are fast becoming defacto standard graphical user interfaces (GUIs) due to the following factors:-
This figure shows the architecture and processes required to provide a Web page based GUI to existing computer applications. We define a legacy system as any generic computer/information system to which we can provide a Web based interface. This includes traditional legacy databases, as well as new systems, such as office administration tools.
The user (1) requests a web page (2) which interacts with a legacy system. This request is transmitted to the external web server over the internet (3). Relevant data is extracted via CGI scripts and sent across the firewall to the application server, where a script command file is executed (4). This reformats the data and launches the legacy application request (5). The response (6) is passed through a dynamic page generation facility and HTTP (7) is returned across the firewall to the external server. The new page (2,) is downloaded via the internet (3) to the user's Web browser (1) and the transaction is completed.
It must be noted that the security issues raised in this paper are not unique to Web/legacy integrations. Therefore, use of existing and well tested security models can be readily employed.
The first area, data protection, is essentially Web server security. There is considerable server security information [HREF 4] available, which discusses ways of improving security.
The security technique
most relevant to commercial use of Web pages for legacy system access,
restriction. This includes restriction by IP address, time domain
and user name/password. Typically a server can be setup via
configuration files (e.g.
[HREF 5] ) to restrict
access to certain machines to certain users.
These methods can help to prevent attacks on the server itself. The second issue is data security over the Internet. The HTTP protocol sends all data unencrypted over the Internet, meaning that passwords and other sensitive data (e.g., credit card numbers) are open to attacks from packet sniffers.
There exists a "Flock of birds" mentality, i.e., there are many eyes checking that no trouble occurs. If something does occur, then it is very likely that it will happen to someone else.
For commercial systems, this approach is too lax. A better method is to use data encryption. Secure HTTP ( SHTTP or HTTPS in URLs) allows all data sent over the internet to be encrypted. As each session uses a new key, a 'hacked' key is only good until the session closes. SHTTP is based on the Secure Socket Layer protocol (SSL) [HREF 6] . Other methods for secure transactions include SHEN and Secure Transaction Technology (STT).
Servers which support SHTTP include - Commerce Builder, Netscape Commerce, OpenMkt Secure WS and WebSTAR [HREF 7] .
The third area of data integrity is the ability to recover after an erroneous transaction. Currently the log files of the Web server daemon can record the name, date and web page accessed. However, this is not enough information for site transaction management. Additional project specific logging is required to record important user data fields and transaction record numbers.
Much research and development has been made into the area of legacy system security (e.g., relational database management systems provide well tested security and transaction management). Whilst highly relevant, these methods of transaction recovery are outside the scope of this paper.
Secure transport methods, such as SHTTP use public key cryptography methods, which raise the issues of key handling and trust, i.e., your public key must be available, but you don't want it compromised. Security certificates provide a mechanism for increasing the level of trust.
Security certificates are a digitally signed certificate including the user's name and public key. Digital signatures ensure the security of the information by appending an encrypted summary of the message to the message itself. Thus if one byte of the message is changed, then the whole message is invalidated.
These certificates can easily be created using mathematical functions which can lead to security loop holes. Thus to increase the level of trust, Certification Authorities, such as RSA, provide a certificate signing service. These certificates include the Certification Authority's own digital signature which cannot readily be forged. To protect the clients, often browsers such as Netscape only allow connections to secure servers which have certificates signed by one of the recognised Certification Authorities.
To obtain a security certificate an application must be lodged with a Certification Authority. This application will include personal details such as name and organisation. Once this application has been verified, a digitally signed certificate is issued.
For computers inside the firewall who need access to external Web sites, a method to pass through the firewall is required. One method is not to filter HTTP messages. However, as HTTP includes services such as ftp, this can be dangerous.
A better method is to use an HTTP proxy server, sitting on the firewall. This proxy can forward HTTP messages between the internal network and the external server.
The CERN HTTPD [HREF 9] server can act as such a proxy, but it doesn't allow SHTTP requests to be forwarded. However, see [HREF 7] for servers which can be used as a SHTTP proxy and will then forward SHTTP messages.
Additional functionality required would include web page & server management utilities. Currently tools are relatively undeveloped in this area and no means of monitoring web servers via existing management platforms exist. First generation tools are now beginning to appear, but functionality varies between server platform.
Problems with a legacy system rather than the Web server will remain difficult to detect from the server unless specific monitoring or integration utilities are developed (e.g. via a Java applet).
For sophisticated user interfaces, a mix of static and dynamic web page generation is required. This will allow the server to generate the user interface page (web form) at the time of the page request. A dynamically generated form can include very specific default data and/or optional fields (such as date, user name, and organisation), allowing different functionality to be exposed on the page, depending upon the requester. Performance of dynamic page generation systems currently is not an issue, as available bandwidth is the current bottleneck, and will remain so in the short term.
One advantage of dynamic web page generation systems, is that configuration management is simplified. This is because as soon as the page generation code has been updated, no HTML generation or directory management is required. Instead, when the page is next requested, the updates will be automatically included when the page is downloaded to the user's browser.
A public interface can also be regarded as a restricted user interface, offering the lowest functionality. This will probably be complimented with additional marketing and pricing data and can be closely integrated with the existing public (external) web pages of an organisation.
The user requires a meaningful response even if the legacy system itself fails to respond. Consequently a timeout mechanism is required which traps legacy system failure. If the timeout period elapses an appropriate error message is returned to the user. Any delayed response from the legacy system will need to be handled by the server, with possible cancellation of the transaction.
If the application returns an error, then this must be passed on/expanded for the user, so that a correctly formatted query can be made. Most non-trivial applications will generate a wide variety of error messages, each of which may result in a different response for the user.
Another legacy integration technique used is screen scraping. Here the web server interfaces to a process which acts as a virtual terminal on the existing system. The web page user-entered data is converted to legacy application commands. Returned data is extracted and formatted into HTML before being displayed on the browser.
However, even with well defined APIs, CGI programming requires significant effort to achieve legacy system integration. Tools are essential for timely and economic development of such systems.
There are a multitude of problems facing a Web administrator. Often a sense of ordering is inherent in the pages, which leads to the concept of forward and backward links - i.e., go to the next/previous page. Initially, these are easy to implement, however difficulties arise if pages are moved, inserted or deleted.
Further, the issue of standard page design and non-functional pages must be addressed. Certainly both the content and look and feel of a page are important. To have a professional and complete aura at a site, all pages benefit from a common layout i.e., same font size, background and foreground colours.
Additional features, such as company logo, common tool bars (at the top and bottom of pages) for navigation around the site and to external links, last updated field, mail to, etc., all prove useful.
These concepts of common layout are relatively easy to adhere to when only a handful of pages are required. Yet a typical site will have hundreds or thousands of pages. When these are generated by different developers and often dynamically, as a result of CGI scripts, common layout is difficult to enforce.
Often this results in pages which are non functional and boring, which leads to user frustration. This robs the site of impact and often basic hypertext functionality is not fully utilised.
A more exasperating situation for users is broken links. These are links to files which no longer exist. This can happen for a variety of reasons, such as file system or page reorganisation. Broken links reduce the user's confidence in a site and hence the appeal of the site.
Simpler internal enterprise solutions involve database searching and email integration (for reservations and loan advise etc.). These applications do not necessarily integrate with existing legacy systems, although flat file searching may be integrated very easily. A public library system [HREF 11] has been implemented in New Zealand, and this demonstrates how the basic 'legacy' functions canbe augmented with extra internet functionality.
Web based applications can be developed cost effectively which avoid these problems. Further, for applications local to the office, (i.e.) intranet, security is less of an issue as employees can usually be trusted. Expect a number of in-house, public domain or cheap solutions to become available shortly!
This paper describes the processes and some of the issues, which CiTR encountered, when we developed an web front end to existing application software. The lessons learnt in developing suites of web pages, the need for security, ease of management and integration aids focussed our attention on the desirability of tools. Consequently which we have developed for use in future integration work.
|Pointers to Abstract and Conference Presentation|
|Abstract||Conference Presentation||Papers & posters in this theme||All Papers & posters||AusWeb96 Home Page|