I'm not sure how the canonical system works.

Will anyone ever link to your canonical URL if it doesn't actually exist? Do the search engines consider IBLs to non-canonical URLs equivalent to links to the canonical URL?

Hello Skuba.

Whenever you have a product type or category type that offers any kind of variations, and those variations are expressed as new pages, the canonical tag should be used.

Additionally, if multiple pages are generated to show products in one category or one explicit type, and the result is a paginated list, the pages should be tagged.

I always use the original page as the page to reference for the canonical link tag. For example, if I have the following

/my-product/
/my-product/blue
/my-product/green
/my-product/red

Then I would setup the link canonical tag to point to the full url + /my-product/ (the other pages are only color variations)

Hope this helps

If you can avoid using the canonical tag - that is the best option.

If your website has pages that are identical or VERY similar (i.e. duplicate content issues) then you should really focus on removing the duplicate content. If you are using URL-parameters (to track user actions, session IDs etc.) then the optimal solution is to do that differently. Or if you are serving the same content for similar terms in product or category pages... the same applies

We have invested heavily in removing dupes and setting up 301s to the correct pages.

If your site architecture doesn't allow for the optimal solution then using canonical AND noindex tags is the next best choice.

The canonical "leaks" more linkjuice than a 301 (which will pass close to 90% of anchor and link power). Since the canonical is only a suggestion, I would always cover my bases with the noindex to avoid duplicates.

You should also try your very best to make sure bots don't crawl the duplicate URLs unnecessarily. That is to say, make sure links on your page are clean. If users decide to link to a page and they copy the "duplicate" URL - this is where the noindex and canonical will make sure that when the bot follows the copied URL link most (but not all) of the link power is passed to the correct page.

Those pages I mentioned are not duplicates. I do want them to get indexed.
Thanks

What I am trying to figure out is if in the case of a COLOR drill down or a results per page change, should I use the canonical tag to link to the parent level like I suggested on my original post?
Thanks

You're describing "faceted navigation" which can rapidly grow the number of URLs available on your domain. You often need a lot of backlinks before Google will even think about a deep indexing like that. I even know a large site with a PR8 home page that gave up on getting all their faceted URLs indexed.

IMO you are better off aiming to get some higher level "hub pages" indexed.

[edited by: tedster at 7:26 am (utc) on Apr 16, 2010]

i'm with what tedster is saying.

however - to respond to the specific question:

in the case of a COLOR drill down or a results per page change, should I use the canonical tag

Just to make sure i am getting this right you have some sort of "search filter". I am going to say you sell shoes (because i need an example). If I search your page for tennis shoes i will probably land on example.com/tennis-shoes/
If I understand you correctly you send me to a new page if I only want to see red tennis shoes. Here i would not use the canonical tag. I would however probably set it to "noindex" depending on how many backlinks you have and how well you are indexed. You could always remove the noindex tag for a small subseciton and test what the effect is.
In the next example you gave us where you have /show-all and /show24perpage etc. I would use the canonical tag AND the noindex tag, because that will definately trigger a duplicate content issue. You are serving identical information for:
parts of /tennis-shoes/
and all of /tennis-shoes/red AND /tennis-shoes/red/show-all AND all other show# pages.