Dashes vs. underscores

Interestingly enough, the example from the post doesn't work anymore:

http://www.google.com/search?hl=en&q=_maxint

Gives "Showing results for MAXINT. Search instead for _MAXINT." Clicking on the second link still searches for MAXINT. You have to use +_MAXINT as the search query for the search to work (and it's not even case-sensitive).

Maybe Google treats underscores the same way as hyphens now that it's mainstream and the majority of people don't really care about the semantic distinction between the two? The post is 6 years old after all.

This article is interesting, but it's really old (2005). There has been discussion about it since then.

Here is an article from 2007 on Matt's blog: http://www.mattcutts.com/blog/whitehat-seo-tips-for-bloggers... It says, at that time, Google was looking into making underscores word separators.

Later, in 2009, Matt mentioned in a Google Webmaster Help video that he'd still recommend dashes if you can, but don't worry about it if underscores are working for you. They might still get around to making underscores separators: http://www.youtube.com/watch?v=Q3SFVfDIS5k

As of 6/20/2011, Google still recommended using Dashes in URLs here: http://www.google.com/support/webmasters/bin/answer.py?answe...

I wonder if this is still accurate. Google has changed a huge amount in the past 6 years. These days, I notice a lot of important punctuation (along with entire words, intentional spellings of words, etc) are ignored by Google in search strings.

The reason not to use underscores is because people don't understand them. I learned this the hard way back in the late nineties the first time I set up an email account for a business. I chose yyy_architects@emailhost.com instead of using a dash because of the connotations the underscore carried regarding computer programming conventions. Then I enjoyed the opportunity to explain over the phone exactly which key was required to correctly type our email address many many times. [Stylized conversation]

   It's an underscore not a dash
   Huh?
   It's the the same key as the dash, only hold shift.
   Huh?
   See the dash key?
   Yeh.
   Just hold shift when you press it.
   Huh?
   Do you have our fax number?

That's why you don't use underscores in a URL. Of course, today nobody uses dashes either if they can help it.

tl;dr:

dashes separate words, underscore is word character -- in common interpretation by, for example, Google Search.

When implementing any text search or matching, there's matter of what constitues word boundary. Obviously whitespace does, as do some punctuation; but exactly which punctuation should be considered word boundary? There is an document on the subject by Unicode Consortium: http://www.unicode.org/reports/tr29/

There is one important case where punctuation and whitespace do not constitue word boundary: the hyphen, as used for breaking word in hyphenation. In Unicode, U+2010 Hyphen, U+2011 Non-breaking hyphen and U+002D (Hyphen-minus, the usual `minus' character from keyboard, for backward compatibility with ASCII / ISO8859-*)

Article dated August 25th, 2005.

I think the bigger problem is that most people don't know that dashes (or underscores) are even allowed in domain names. Thus we get domains like "penisland.com" for a company called Pen Island. Short of that kind of problem, multi-word names with no dashes are just hard to read.

So remember, the next time you go to register a multi-word domain name, use dashes!

I guess Google didn't have any lisp programmers in the early days!

There have been articles, on Google's Analytics-Support page(s), that have recommended otherwise or have made not that underscores are just the same. Some of those old links no longer work -- answer (id) expired, but the following still has language in support of underscores: http://www.google.com/support/googleanalytics/bin/answer.py?...

I was hoping this would be serious study on whether dash-style or underscore_style is better for variable names in code. Such a study would analyze readability, ease of typing, and any other relevant factors. Too bad this article isn’t such a study, while the information it does impart is out-of-date.

There's been massive advancements and updates since this was written in 2005 so don't expect this to be gospel.

The answer is the article URL.