- From: James Graham <jg307@cam.ac.uk>
- Date: Wed, 16 Apr 2008 12:41:02 +0100
- To: Steven Faulkner <faulkner.steve@gmail.com>
- CC: public-html@w3.org, wai-xtech@w3.org, wai-liaison@w3.org
Steven Faulkner wrote: > Take the example of the john's flickr page cited by Ian earlier, that > page contained 24 images without an alt attribute. There is no > reliable means to determine whether any of these images contain > information important enough "critical to understanding the page" to > convey their presence to the user. If all our conveyed in some way, > the user would hear the word "graphic" (for example) 24 times > sprinkled throughout the text content of the page, adding to the > cognitive load on the user without aiding understanding. To get any > information from these images, the user would have to set "virtual" > focus to each one in turn (for example in JAWS, users can navigate > from image to image using the G key) and then use a series of > keystrokes to query the attribute values (src for example). In most > cases the attributes will not contain any useful information, so would > be an exercise in futiltity for the user. As well as taking a very > long time. > > does that help? Yeah, I see the problem with the scenario that you mention. There are several things worth noting: The optimum behavior with different classes of UA may be different. e.g. in a text-only browser providing some information for graphics with no alt attribute is typically reasonable and useful. The problem you mention may not go away with a noalt attribute. It seems pretty likely to me that the attribute would often be misused to label decorative images as not having any alternate text, leading to the same usability problems you describe. Indeed the problem exists even when an explicit but poor value is chosen for the alt attribute e.g. I was sent an html email that reads as: "Oxford Playhouse Logo Floating Kate Rusby Jeremy Hardy Floating Kate Rusby Jeremy Hardy Floating Kate Rusby Jeremy Hardy Floating Tuesday 11 September [...]" with images disabled but reads much more sensibly in the text/plain version: "The Oxford Playhouse What's On View Online http://www.oxfordplayhouse.com/eimages/mails/10227. htm ********************************************************************** ******************************************************* Tuesday 11 September [...]" It seems possible that the problem could be fixed with a better heuristics and a better UI for the AT. For example the AT could ignore images with common banner shapes, it could provide a power user option to ignore all images in a certain domain/path (this would work for flickr, for example), or could provide more sophisticated analysis of the image contents to distinguish e.g. icons from photos and to identify features in photos (these techniques are becoming increasingly common in digital cameras, for example). Depending on the results of the analysis the graphic should be presented at lower/higher verbosity levels. -- "Eternity's a terrible thought. I mean, where's it all going to end?" -- Tom Stoppard, Rosencrantz and Guildenstern are Dead
Received on Wednesday, 16 April 2008 11:42:42 UTC