The … used when truncating the question body in the title attribute of links on the front page is being double-encoded, as seen below:
Additionally, the space before the … should likely not be there. There's also an extra space after it as well, but I'm not sure if that's noticeable without looking at the source.
This derives from the subtle difference between HTML Encoding, and Attribute Encoding.
We don't actually reach in to revision history to get the raw plain text of the Markdown for performance reasons, so we always start with the HTML version of the post... which kind of works for attributes, but HTML Encoding != Attribute Encoding.
Adding a new method to synthesize explicit plain text summaries here by (hack) using HtmlDecode rather than the previous "HTML with the tags removed" method, which isn't really correct.
