There are two ways to view most sites: the www and non-www version. This has led to problems with search engines indexing both versions, resulting in duplicate content and diluting your inbound links between the two domains.
Search engines are mostly able to determine whether the www and non-www versions are the same and combine the results (passing all the link benefit to one version). However, sometimes this fails and we see both versions indexed, known as a www/non-www canonical issue. This is bad for the site owner since link benefit (Page Rank/anchor text) is shared over two sites.
It's good SEO practice to resolve canonical issues because they can impact your site's ranking. Here's a good article on the subject.
In addition to the sub-domain level canonicalisation issues set out above, sometimes content management systems and e-commerce systems will spew out lots of duplicate content pages which also dilute your SEO efforts. Google now supports the rel=canonical tag, by adding this tag within the <head> section of any pages with duplicate content you can tell the search engines which is the page that you want them to rank. You can read more about this on the Google Webmaster Blog.