PS to Mark: I don't think you do have to support dc:date. Dates are not absolutely required to deal with a RSS feed. After all at the item level 0.91 didn't have them at all, and a lot of feeds are in that format. If an aggregator is going to work with those feeds it has to do something meaningful with items with no pubDates.

Brad: People weren't studying the spec, as you say you are doing now. That's what I realized when I looked at how otherwise intelligent people were mangling RSS. Not just well-formed manglings, but (for example) 0.91 feeds with guids (that's not cool, guid didn't come in until 2.0).

Mark: Prior art is everything. The rule for evolution from 0.91 to 0.92 to 2.0 is that everything new is optional and if you put a 2.0 version number on a 0.91 file you get a valid 2.0 file.

Anyway, it feels like this discussion is going somewhere. Happy about that.

Mark: Did I mention RSS 1.0 at all? My hypothetical RSS 2.0 reader was built to just read RSS 2.0 files. Without consideration of namespaces, it wouldn't be able to date entries unless pubDate is present in the feed. In such a case, I would consider that a failing in the RSS 2.0 feed, not in the reader.

Dave: I admit I've been largely oblivious to all this until recently. I've heard bits and pieces of the discussions (I would guess a lot of webloggers fall into this category).

dcterms:modified uses the 8601 format like dc:date. I find that the HTTP 1.1 Last-Modified is more useful then the dcterms:modified date simply because its easier to obtain even if its not totally accurate.