Should aggregators parse relative URLs?

Question: should RSS aggregators and newsreaders parse (expand) relative URLs in RSS items?

My thinking is no, they shouldn’t. As I see it, the idea behind syndication is transportability. The content of an RSS item is not relative to a location on the Web. (Which is different from HTML pages, for instance, which are relative to their own URLs.)

But nevertheless I get bug reports—people tell me images don’t show up for a certain feed, or the URLs don’t work. And it’s because they’re relative URLs.

What do you think?

I can, of course, see the other point of view also (that RSS feeds are relative to the link specified in the channel header, usually the home page of the site).

I want to do what’s right, but I’m not sure which is right.