I appreciate all the comments and points of view about this, and I still think it's a workable idea (even if given another label like "legend") and that it ought to be carefully considered for inclusion in one form or another.
> How is your proposed CAPTION attrbute different from the TITLE attribute > already included in HTML4? Alexey, the only problem I have with "title" in this context is in the current UA implementations. I'm not as familiar with the W3C specs as many people here, but in practice "title" doesn't display, does it? In Firefox, at least, it displays when the cursor hovers over it. I don't want to explain to non-savvy users that they need to hover over images to get the caption. Instead, we should have a simple way to attach a caption that all user agents interpret as being associated with a particular image. The visual display would be controlled with CSS, but its context should be readily understood by JAWS, Window Eyes, and other non-graphical user agents. I should explain that my own first priority is to serve accessible content to all users. I don't want to create "hacks" to accomodate blind users any more than I want to create "hacks" for non-compliant user agents. There's been too much of that already, and it needs to stop. Since Gutenberg, we've had this convention of the caption that associates directly with a graphic image. I think we can do this in HTML/XHTML in a way that is simple and consistent, semantically correct and robust enough to meet most needs. > The problem is that captions can and do have substructure. For instance, > a caption might include multiple emphasized or strongly emphasized > sections. Attributes just aren't powerful enough for this. Why is that, Elliotte? Maybe I just expect less of a caption. After all, don't we just need it to provide an explanation for why or how this graphic image relates to content? It's the content that should carry the real burden. A caption may fulfill a more elaborate purpose in technical and scientific papers, but I don't see that as inherently necessary. I'm open to understand differently, though. Anyway, I guess one of the sticky problems here is in the need to "degrade gracefully." Older UAs would ignore a new attribute. Many of my users will still be left out until their UAs are updated to reflect whatever standard does evolve to handle this appropriately. Unlike graphical browsers, their user agents are not available for free download.
<<winmail.dat>>
