Diagnostics
One will not accept the advice of a doctor unless he realizes he is sick. This same principle applies directly to the state of website design, and in keeping with the metaphor, the next phase of research is to diagnose just how inaccessible well-known websites currently are. With this goal in mind, I developed several methods for analyzing the accessibility and compliance, selected forty sites to diagnose, and compiled the results to see what kind of conclusions can be drawn.
Disclaimer:
The tests and analysis done on these example sites are intended to serve as the basis for discussion rather than to illustrate either effective or ineffective design choices or technical administration.
The Patients and the Tools
List of Test Websites: The forty subjects organized by category, including links to the pages tested and brief commentary on why each was selected.
The Tests
HTML/XHTML Validation Statistics: For each site, I measured how many HTML 4.0 and XHTML 1.0 Strict validation errors it triggered and then compared those numbers to the number of lines of code of HTML to determine the ratio of errors-to-code.
Section 508: For this test, I parsed each test site using the Bobby software to measure the site against the Section 508 accessibility guidelines. I recorded the number of errors, user checks, and lines of code and then computed the percentage of the number of lines of code that contain errors.
Web Content Accessibility Guidelines 1.0: Similar to the Section 508 test, I used Bobby to measure each site against the WCAG accessibility standards published by the World Wide Web Consortium. I recorded the number of errors under Priorities 1, 2, and 3, and I computed the percentage of the number of lines of code that contain errors.
Code Ratio: This test measures the size of each site's HTML code and the size of the actual textual content on the page. Comparing these numbers gives us information about how "bloated" the code may be for sites that use a lot of extraneous markup to render the page. We then can determine the correlation between low code ratios and "accessibility."
Analysis of Results
HTML Validation Results
Number of errors per line of code (average)
- Popular sites: 1.37
- UNC-CH related: 0.85
- Search engines: 8.80
- Government: 0.86
- Accessibility Advocates: 0.18
Section 508 Compliance
Percent of lines of code with errors (average)
- Popular sites: 7.08 %
- UNC-CH related: 3.92 %
- Search engines: 18.28 %
- Government: 1.12 %
- Accessibility Advocates: 0.14 %
WCAG 1.0 Results
Percent of lines of code with errors (average)
- Popular sites: 31.60 %
- UNC-CH related: 20.73 %
- Search engines: 71.56 %
- Government: 36.99 %
- Accessibility Advocates: 6.51 %
Code Ratio
Bytes of HTML code / Bytes of Textual Content (average)
- Popular sites: 11.28
- UNC-CH related: 10.38
- Search engines: 7.49
- Government: 8.63
- Accessibility Advocates: 3.15
Conclusions
By all four metrics I used to test these sites, the content providers who support web standards an accessibility far outshone the other genres of websites. The Popular and Search Engine sites in particular fared very poorly in all validation measures, which is a major concern given their widespread appeal to users of all types, including those with disabilities. These entities are also those that are covered by the ADA, which raises the question of litigation -- and AOL itself is proof that it lawsuits can and do occur.
The Code Ratio test highlights a secondary problem for which universal Web design is a solution. As shown in this diagnosis, the sites that are the most inaccessible by official standards are also the most bloated, in that they require a whole lot of HTML to put their content on the screen. For the popular sites, it takes nearly 12 characters of HTML to generate one character of actual text that, say, a screen reader can process*. On the other hand, the ratio is nearly 3:1 for the accessible sites. This indicates that there are major bandwidth implications of inaccessible sites above and beyond the accessibility issue. If a site like ABC News is queried hundreds of thousands of times a day and can drop its code ratio from its current 20 to 3-5 using accessibility techniques, the traffic load on the Internet would drop drastically. Multiply this effect across the hundreds of the most popular sites on the Web, and the potential reduction of bandwidth utilization would be ridiculous.
* Of course, there may be some content embedded in the <alt> text that is not included in these calculations, but it is assumed the net overall effect of including <alt> text would not be significant (and the costs of doing so are too high).
