I know from nothing about RSS. But, here's a truncated version of this site's RSS source:
<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0">
<channel>
<title>Interrobang Cartel message board</title>
<link>
http://www.interrobangcartel.com/forums/index.html</link> <description>Posts on Interrobang Cartel message board at www.interrobangcartel.com.</description>
<item>
<title>RSS (2)</title>
<link>
http://www.interrobangcartel.com/forums/kareha.pl/1218451990/</link> <guid>
http://www.interrobangcartel.com/forums/kareha.pl/1218451990/</guid> <comments>
http://www.interrobangcartel.com/forums/kareha.pl/1218451990/</comments> <author>Doctroid</author>
<description><![CDATA[ <p>What's with this board's RSS? Goolge Reader doesn't see half the new posts. But I just tried using Safari's builtin RSS, and it does seem to get the new posts. Unfortunately I normally do not use Safari.</p><p>Is there something nonstandard about the RSS source?</p> ]]></description>
</item>
[...]
</channel>
</rss>
And here's RSS source from Ravelry, which as far as I can tell is handled correctly by Goolge Reader (as are, to my knowledge, all the other non-IC feeds I subscribe to, but then again I didn't notice the problem with this feed for quite a while):
<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:dc="
http://purl.org/dc/elements/1.1/"> <channel>
<title>Doctroid's Ravelry friend activity (filtered: project photo, stash photo, queue, fave, magic link, comment, handspun)</title>
<link>
http://www.ravelry.com/people/Doctroid/friends/activity?</link> <pubDate>Sun, 10 Aug 2008 01:26:55 GMT</pubDate>
<description>Doctroid's Ravelry friend activity (filtered: project photo, stash photo, queue, fave, magic link, comment, handspun)</description>
<item>
<title>jwgh's Cube blanket</title>
<link>
http://www.ravelry.com/projects/jwgh/fibo-optic</link> <description><p class='user'><a href="
http://www.ravelry.com/people/jwgh"><img border="1" src="
http://avatars.ravelry.com/jwgh/68853/1115660-1_medium.jpg" style="border: 1px solid #666666;" /></a></p><p class='type'><img class="icon activity_icon" height="16" src="
http://assets.ravelry.com/images/silk-color_swatch.png" width="16" /> <a href="
http://www.ravelry.com/projects/jwgh/fibo-optic">jwgh's Cube blanket</a> </p><p class='photo'><img border="1" src="
http://farm4.static.flickr.com/3242/2744477157_bb31d0bba6.jpg" style="border: 1px solid #666666;" /></p><div class="notes" style="margin-top:1em;"></div></description>
<pubDate>Fri, 08 Aug 2008 22:01:47 GMT</pubDate>
<guid>
http://www.ravelry.com/projects/jwgh/fibo-optic?activity=8251608</guid> <author>jwgh</author>
</item>
[...]
</channel>
</rss>
The elements are in different order, but I assume that's not significant. More substantive-looking differences are:
(1) Ravelry's rss element includes the xmlns:dc attribute; IC's doesn't.
(2) Ravelry's channels and items both have pubDate elements; IC's doesn't. (But the data are human readable so I suspect are not parsed by the reader software.)
(3) Ravelry's item description is some HTML with the < and > turned into < and >; IC's is the ![CDATA[ tag.
The IC item above is one that was shown by Goolge, but it doesn't look to be substantively different from items that weren't.
How does the reader software know what's been read already and what hasn't? That is, when it gets the RSS source, how does it decided which items to show? It seems to me after a new thread is started, any time that thread is updated the only change in the RSS will be in the item title, which shows a larger message count than before. Since there is no pubDate element it can't use that to decide, and since the description element is based on the start of the thread rather than the new message in the thread (which, by the way, is annoying), it can't use that either. Hmm, if I look at all IC items in Goolge Reader, it shows "All Roads Lead Away From Rome" twice, once with a post count of 18 (in the title) dated last Dec 6, and one with a post count of 11 dated Dec 13! But in fact posts 11 and 18 were made yesterday and today, respectively. Weird, weird.
OK, I just hit refresh and the entry dated Dec 6 now shows the post count as 20. So it's seeing the IC entry in the RSS but deciding it's something I've already read, even though the title has changed.
... OK, just checked some other feeds; most use pubDate elements but some don't, so that doesn't seem to be relevant. And some use ![CDATA[, so that's probably not it either.
So could it be Google Reader is simply regarding the item as unchanged and hence read because its description is unchanged? But then, how come it shows two 'All Roads' items, having the same description? Weird.