Rogers Cadenhead: This sounds good, though it officially abandons the pretense that Google's search algorithm is tailored to the linking behavior of Web users, rather than the other way around.
The lines between us and them is blurry. Everything is intertwingled. Whether we admitted it before or not.
My urls contain dashes instead of underscores because
google likes them better. My <title>
s
are meaningful for similar reasons.
In return, I get a more traffic from Google.
One of the things I like best about nofollow is that it is guaranteed to make search engine results less full of crap. It may not reduce the amount of comment spam hitting weblogs but it definitely will reduce the number of artificially highly ranked weblogs and comment spammed sites that show up as a result of searching Google, MSN or Yahoo!.
The fact that the link to my blog on this comment no longer gets high PageRank from being linked to from Sam Ruby's blog is a small price to pay IMHO.
PS: You should add MSN, PageRank and nofollow to your dictionary.
Hm, I don't think the method proving that Google likes dashes more than underscores (that you link to) makes sense.
He's testing the search term parser (from the web search form) and using that to explain how the Google bot parses URLs, right?
In any case, I agree that what we do is (and long has been) affected by what Google and other search engines do. But, I don't agree that nofollow will make search results better (details here) or lessen comment spam.
The article is right, but more than a little muddy in explaining itself. Since you rarely see keywords in a URL that aren't also in the body of the page, the easiest way to test it is by doing an allinurl: search:
allinurl:follow up +on nofollow
allinurl:underscores +are bad mmkay
allinurl:underscores_are_bad_mmkay
The first shows that Google believes all four words appear in the URL for this post, the second shows it doesn't believe all four separate words appear in the URL for the post when I first heard about this, and the third shows it believes there's just one word, a word which includes underscores, in that URL.
Rumor has it that they don't treat underscores as word delimiters as a favor for programmers, to have decode_utf8_string only find a function name in code, not a page which includes all three words.
Thanks Phil--that's pretty convincing.
I guess it's still imaginable that the Google allinurl parser function is different than the Google bot URL parser, e.g., the reason you cite is useful for human query construction but not necessarily for analysis of terms on the page.