Implementation Models

There are a number of different approaches to providing access features for Web content.  In order to facilitate discussion we have categorized them into four basic implementation models:

Model 1: Downloaded user agents

Applications created under Model 1 are designed to be downloaded and installed on the user's computer. The FireVox and LowBrowse plug-ins to Mozilla Firefox are examples of this model. AT that runs in combination with browsers would be another example. The IBM Home Page Reader would be an older example of this approach.

This model takes advantage of the client computer's processing power in order to create faster response than server-based solutions. Additionally, these types of projects are currently the easiest to develop and deploy (especially the browser add-ons), since they require no added server side infrastructure. However, they require that users have permission to install and modify programs on every computer on which they may need access.

NOTE: We are working on a special tool that does allow some downloadable agents to be run on a computer without installation to the disk.  This approach is under development and works on some operating systems.  It will not likely work on a heavily locked down system however without cooperation from the system. 

Model 2: On-Demand Web Services

Model 2 represents web services that provide transformations on demand. DocAccess (which converts page images to text on demand) is an example of this approach.

These services may be called automatically from content, by a user, or by other web services. The web services will perform conversions on web content that is sent to them; for example, they may translate images of text into accessible text, shift colors to accomodate a specific type of colorblindness, or convert text into audio. Because these services are hosted on a remote server, they have access to processing power beyond that of the host machine. They can also be used without any downloads or installation. 

Model 3: Proxy-based transcoding

Model 3 projects are developing proxy services that interpret and re-present content on its way to the user.

These solutions require some configuration by the user to allow content to pass through the proxy gateway, so the user must have permission to modify internet settings on his computer. After initial setup, however, the user may browse normally and automatically receive accessible pages as modified by the proxy.  There are limitations however regarding secure socket layer (SSL) connections.  There have also been copyright issues raised in connection with this approach. 

Model 4: Web-based user agents

Model 4 projects include web-based access technologies that can be used anywhere, from any internet-enabled device.  WebAnywhere is an example of this approach.

With this approach processing is done at a remote server, accessed through a specified URL, and does not involve any downloads or installation at the user's computer. Because these solutions reside on the server instead of the user's computer, personalized settings can be retrieved wherever the user logs in. This approach is particularly effective for individuals without resources who often are accessing the web through other people's computers, libraries, or community centers. 

Hybrids

It is possible to have solutions that involve aspects of all these approaches.  This can take advantage of the strengths of the different approaches will avoiding some of their limitations.  The 4 Models are useful in thinking about and discussing the projects and so they are used throughout this site.