Source code was checked with A-Prompt to verify accessibility guidelines that might not be determinable via browsers. A-Prompt searches the source code file against the W3C/WAI Guidelines and gives options for an immediate fix of the problem. If a fix is selected, A-Prompt inserts the correct code into the file. A-Prompt reports items that may not actually be a problem. For instance, it reported that Blackboard 4 contains images that are missing descriptive text, even though the image was not complex enough to need it.
Our final validation of accessibility were the guideline checklists, the W3C/WAI guidelines checklist (or, our easy-to-print W3C/WAI checklist) and the Section 508 checklist (Word). The W3C/WAI checklist is an easy-to-follow list of the guidelines ordered by priority and broken down into subcategories such as table use and form use. A "yes", "no", or "not applicable" is given to the series of points. We created the similar checklist for Section 508 for Fall 2001.
The checklists provide a simple yes/no format and a final crosscheck of all accessibility traits. The list was checked across the full scope of pages and was always a consideration. As accessibility problems were found, they were noted on the checklists. If in even one place the program did meet the guideline, the program failed that checkpoint.
Our A-Prompt evaluation highlighted errors that were difficult to determine in the other parts of the review. It also confirmed some errors that we had already found.
It was difficult to evaluate a set of pages that were the same across all programs; instead, we reviewed a sample content page from each program. Content pages are the most important aspect of LMS packages, as these are most likely to contain vital course information.
A-Prompt revealed a number of similarities across the products. The same seven errors occurred in each course package. A-Prompt reported missing DOCTYPEs, fixed font sizes, missing descriptive text for images, link text that is not meaningful, missing captions, and missing summaries for tables.
A-Prompt found unique errors in the products as well. Two products, Blackboard 5 and Prometheus, do not identify the language of the page. All products save Blackboard 4 had JavaScript problems. These products provided no alternative to JavaScript-based functions, such as login and chat forums. This issue is reflected in the reader tests we did with these products. A-Prompt found suspicious alt tags and input label problems in WebCT, and auto-refresh errors in Blackboard 5.
The W3C/WAI guidelines checklist show high error numbers. We found 26 failures in Blackboard 4, six of which were Priority 1. Prometheus yielded 30 failures, four of which were Priority 1. WebCT had 31 failures and Blackboard 5 had 33; both programs had five Priority 1 errors. We evaluated the checklist continuously throughout our testing, and failed a program for one error at any point on any page within the package.
Twelve errors were consistent across all four programs. This includes the lack of alternative content for script and applet pages, not using style sheets where they could, not avoiding pop-up windows, and not describing framesets. None of the programs identifies the primary language of the pages, provides keyboard shortcuts, provides non-link characters between links, or allows users to receive documents according to user preferences.
Consistent with A-Prompt, we found a number of errors unique to a few programs. Blackboard 4 and Prometheus fail to provide text equivalents for non-textual elements. Both versions of Blackboard fail to provide text links for image map content. Blackboard 4 fails to provide alternatives to JavaScript elements and does not title its frames. Most of the packages fail to identify row and column headers in data tables. Blackboard 4 escapes this error because we did not find any examples of data tables in the program.
It is interesting to note that the errors reported by A-Prompt were not always the same as those we found while conducting manual checks. While A-Prompt reported seven errors that were the same across all programs, we found 12 common errors during manual checks. Many of these errors are not the same across the validation tools; in fact, only two errors are common in A-Prompt and the checklist evaluations.
On further analysis, we saw that most of the additional errors found with the manual checks relate to the relationship between pages and the navigational structure of the program itself. Pop-up windows and poor frame structure, for example, are errors that will appear only during a manual check. Validation tools that are checking one page at a time will not catch these errors.
Many of the items reported by A-Prompt are actually manual checks; items such as missing descriptive text, meaningful link text, and caption and summaries for tables are dependent on the context of the images and tables. A-Prompt reports these items as errors, and then requires the user to verify that the errors really do not exist in the page. These errors are conditional to the actual page. Secondary manual checks, as well as W3C/WAI checklist validations, show that none of the images requires descriptive text. In addition, the tables, used for layout of the page, are exempt from captioning and summary requirements.
Content authored by AnnMarie Johnson
and Sean Ruppert. ©2001
last updated
November 20, 2001
by AnnMarie Johnson.