- From: Nir Dagan <nir@nirdagan.com>
- Date: Thu, 17 Dec 1998 17:49:29
- To: w3c-wai-er-ig@w3.org
At 16:14 17/12/98 +0100, Daniel Dardailler wrote: > >> Two Wishes for Bobby (or any accessibility checker) >> >> 1. Condition Bobby approval on validation. >> A document that is invalid should not >> be accessibility-approved. > >Bobby presently has a "trivial" validator which only looks at >element/attribute names, but not content model I think. This is true. But I am afraid that many authors don't realize the difference. Also, you can get Bobby approved even if Bobby finds unknown elements (or whatever it checks for), as they appear as "browser compatibility" which seems to be classified as independent of accessibility. > >It could reuse the W3C HTML validator at http://validator.w3.org, >either by downloading it (source is open) and running it local or by >calling it remotely thru the cgi syntax (we'd need to work together >some kind of stable output syntax for the validator, as it's now only >meant to be parsed by human). > >Given the downloadable nature of Bobby, maybe integrating the W3C >validator locally is better (the drawback being that everytime the >validator is updated, Bobby needs to update its bits). There is another thing that can be implemented independetly: Have in the W3 validator an option to run also in Bobby. Now it is implemented for weblint, a good yet somewhat outdated program. > >> 2. Spell check. (This now in the guidelines) >That's a tough one, spell checking in an international environment >(the web) is already hard, but an international techie one is really >tricky. An option without spell checking, would be that Bobby reports what languages are used in the document and warns about non-marked (language-wise) text. Clearly, without spell checking Bobby can't tell (or hint) on incorrect marking. Another wish: Another thing which should not be rocket science is that the W3C validator and Bobby can render the HTTP response headers to the user (for online documents). I had hard time finding a way to see what my server sends. This is pardoxical as every browser has this info, but most hide it well for some reason. (An exception is Cello, that gives the headers in its "view source" option; but I can't recommend installing Cello just for that; Netscape gives the Content-Type in a non-trivial way in its "view info" option; Opera seems to hide the headers from the user.) http://www.delorie.com/web/headers.html is a service to view the response headers for GET requests. It also has the option of choosing a user-agent string to check for discrimination on this basis. Regards, Nir. Nir Dagan, Ph.D. http://www.nirdagan.com mailto:nir@nirdagan.com "There is nothing quite so practical as a good theory." -- A. Einstein
Received on Thursday, 17 December 1998 11:49:30 UTC