My mother, who has a WordPress blog, keeps telling me about posts that I haven't seen. This was starting to irk me, so I looked into River2 to see what's going on. Yes, it is finding her posts, but it thinks they're pictures. Why? Because the feed says they're pictures. Oy.

Digging in a little deeper. WordPress has a neat feature that I don't fully understand, called "gravatars." If you have one, as my mother does, it attaches it to every post, as an image. I'm sure she has absolutely no idea what a gravatar is, and I'm equally sure that WordPress created one for her automatically.

So, I want to fix this, so her posts (and those of other WordPress authors) show up where they belong in River2. Other than looking at the URL of the image, I have no idea how to do it. I'm hoping one of the readers of this blog does.

Here's an example from the feed.

<media:content url="http://0.gravatar.com/avatar/ed80e40fc7fa8d76b88e3e5d1079f429?s=96&d=identicon&r=G" medium="image">

They use the media:content element to represent the gravatar. I have a strong feeling this is very wrong. It seems to me that a gravatar is a bit of metadata. Why should it be represented as an image, why not as a <gravatar> in a new namespace defined for the purpose of representing gravatars?

The media:content element came into being to help Flickr attach pictures to items in feeds. It probably was a mistake, in hindsight, to try to make a general namespace for this, because it gets us into jams like this. Probably would have worked better if they had come up with a <flickr:picture> element. That way we might not have had this conflict in semantics.

<media:content url="http://farm3.static.flickr.com/2785/4317723874_9ac5cf85e5_o.jpg" type="image/jpeg" height="1880" width="2816"/>

I'm pretty stuck here. I really need to separate pictures from non-picture items (I subscribe to some awesome picture feeds, and they would completely swamp my news-oriented feeds). It looks like I'm going to have to check if the image comes from gravatar.com. That's a terrible way to parse metadata.

 2/1/2010; 9:02:26 AM.   .