A lot of RSS is full of weeds. What do you do?

Danny Ayers' A matter of form post critiques some of Mark Pilgrim's comments in Mark's post about his Universal Feed Parser 3.0 beta release. I added some comments to Danny's post to suggest how it seemed like Danny and Mark were using the word "data" in slightly different ways, which I believe, might account for some of their apparent difference of opinion.

At the end of my comments, I suggested an analogy of ecology and sustainable farming. Later, I was thinking that parsing RSS coming from a variety of sources brings up the issue of dealing with what are the equivalent of weeds in the data.

The purist school is saying something like "don't pick your crops until you are sure there are no weeds: weeds are poisonous and they'll take over, so don't let any get mixed in". (Stop parsing if the data is not well-formed.)

The pragmatist school is saying something like "you need to pick your crops now, and you can remove the weeds as you go: even if weeds are poisonous, you can get enough of them out that no one will get sick and they won't take over". (Keep parsing even if the data is not well-formed: use work-around extraction techniques to get the weeds out.)

My basic sense is that both approaches have their place, and both have negative consequences if followed too strictly too often.