Will rel=canonical break site: queries?

Our company publishes our software product's documentation using a custom-built content management system using a dynamic URL namespace like this:

http://ourproduct.com/documentation/version/pageid

Where "version" is the version number to which the documentation applies, and "pageid" is a unique string which identifies that page in our back-end content management system. For example, if content (e.g. a page about configuration best practices) is unchange开发者_Python百科d from version 3.0 and 4.0 of our product, it'd be reachable by two different URLs:

http://ourproduct.com/documentation/3.0/configuration-best-practices
http://ourproduct.com/documentation/4.0/configuration-best-practices

This URL scheme allows us to scope Google search results to see only documentaiton for a particular product version, like this:

configuration site:ourproduct.com/documentation/4.0

But when the user is searching across all versions, we don't want Google to arbitrarily choose one of the URLs to show in results. Instead, we always want the latest version to show up. Hence our planned use of rel=canonical so we can proscriptively tell Google which URL we want to show up if multiple versions are being searched. (Users who do oddball things like searching 2 versions but not all of them are a corner case, so we don't care which version(s) show up in that case-- the primary use-cases we care about is searching one version or searching all versions)

But what will happen to scoped searches if we do this? If my rel=canonical URL points to version 4.0, but my search is scoped to 3.0, will Google return a result?

Even if you don't know the answer offhand, do you know a site which uses rel=canonical to redirect across folders in a URL namespace. If so, I could run a few Google searches and figure out the answer.

The rel=canonical link element helps search engines to determine the URL that they should index, so ultimately, by specifying it for a URL, you're telling them to drop the old version and only to index the new version. In practice, it might be that both versions are indexed for a while (depending on how they're discovered and crawled), but in the long run only the canonical will generally remain indexed. In other words, if you do this for your site, over time the site:-query results for the old versions will drop (which probably makes sense).

If you need to have both versions indexed, then I wouldn't use the rel=canonical link element, I'd just link from the old versions to the new versions (eg "The current version of this document can be found at X").

Wikia uses rel=canonical link elements fairly extensively, though I don't think they use it in folders, but you can still see the results for individual URLs.