Anne van Kesteren

Image semantics

In this little write-up I will refer to document structures using CSS selectors. If you don't fully understand it let the Selectoracle explain it to you, though I think everyone can follow it easily.

Moose made me think about image semantics when we were discussing object>h1 versus h1>object in a thread about one of his experiments. While I wanted to nest the OBJECT element inside the header element for practical purposes:

But actually, the image itself has semantics. These semantics are not defined by the elements around it, but by the image (maybe the file) itself. A simple search engine can't extract those semantics from the image (I wonder if any software could) and therefore it needs a fallback; are you beginning to see the object>h1 picture?

If you think about this a little longer you are starting to realize that h1>img "doesn't make sense" either (I'm putting quotes around that, this stuff is badly under defined, everyone can have their own interpretation/opinion about it), which leads to another reason to "dislike" the IMG element. We still have to use it though and since there isn't any bot (that I know of) that can read both the semantics of HTML and images, I don't think there is any problem, besides that it is "incorrect".

Jukka K. Korpela wrote in an e-mail to www-style that an image has semantic meaning, like I stated above (After all, a logo (as a specific appearance, specified by image data) is content rather than stylistic variation.) and therefore, it should probably not be expressed using the content property, which is now proposed for CSS3 (my model, my model :-) (note that it may change any time anywhere)). And although he might be correct from a semantic point of view, having the content property gives document authors many options like having multiple style sheets (you might think of a print and screen logo). Personally, I love the CSS method: nl{content:url(nav.svg)}.


  1. A., I am not sure of your opinion after reading this article... Do I understand well that whilst you used to have been against, now you are for? :)

    I hope that this will not be considered pimping, but the link you gave does not direct a user to the experiment itself. I'll be so bold as to link to the experiment synopsis.


    Posted by Moose at

  2. Yes, I agree with you now :-). It just takes a little bit more time.

    Posted by Anne at