Are the SE's pulling out URL's that are NOT embedded links?
Yes, there are mountains of them in gwt error pages-- sometimes even when there
is a link wrapped around the damaged text.
Concrete example under "not found":
hovercraft/h.. That's quoted verbatim, dots and all. If you follow the "linked from" links you arrive eventually at a page with the world's spammiest meta tags and a list of urls, including--
Wait, I've got to do some more verbatim quoting under the vague head of "With friends like these..."
<td width="580">
<div class="msnresult">
<div style="margin-bottom:5px; padding-left: 8px;">
<a href="http://www.example.com/{directory}/{filename}.html" target="_blank" class="msneresult" rel="nofollow">{page title} - {my domain name}</a>
<div class="msnresultcnt">
{text of my meta description}</div><span class="msnresulturl">http://www.example.com/hovercraft/h...</span></div></div></td>
Notice (a) the teeny-weeny detail that the "not found" version snips off one more dot-- truncated urls on the page always have three-- and (b) they seem to have decided that "nofollow" doesn't count on this page. The dot-snipping doesn't kick in after a fixed number of characters, though it may be some physical lenghth in pixels. I ran out of interest at this point ;)
What's notable is that the "real" link, unsnipped, is only two lines away. But that one doesn't count as a link. (I checked a different gwt page.)
Someone, somewhere, programmed a computer to make these decisions.
