you appear to be doing the right things regarding domain canonicalization.
if you weren't, then comparing the sites in GWT might be informative.

if google can resolve the verification file for both hostnames (a simple test) it is assumed you also have control of the non-www hostname.

this current thread in the Sitemaps, Meta Data, and robots.txt forum describes how the crawl rates and number of indexed urls differ between the www and non-www hostnames of the OP's site.
Googlebot not respecting custom crawl rate:
http://www.webmasterworld.com/robots_txt/4325733.htm [webmasterworld.com]

Yes. They are two separate sites, with separate reports.

The 'incoming links' report is most informative as one lists links pointing at the www version and the other lists links pointing at the non-www version.

See too, the crawl report. You can see if Google is taking notice of the non-www to www redirect that should be in place. Make sure that redirect works for all pages not just the root.

[edited by: g1smd at 7:33 am (utc) on Jun 18, 2011]

In WMT settings, if you choose "Don't set a preferred domain" then you wouldn't get this.

if google can resolve the verification file for both hostnames (a simple test) it is assumed you also have control of the non-www hostname.

The verification atm is the verify-v1 tag in the index file, i.e. the HTML at the root; i.e. the headers above show the 301 working for this.

I might add Google Analytics connection as an additional verification method, but don't see it will make any difference.

Yes. They are two separate sites, with separate reports.

Except they're not, and I don't want separate reports (or, more importantly, Google confused about the domain and any negative consequences of that confusion).

I haven't had this issue with other domains with more conventional TLDs but identical set-ups.

Make sure that redirect works for all pages not just the root.

It does, for sure.

Just to check Google's not seeing anything non-canonical, I get zero results in Google for site:example.net.nz -site:www.example.net.nz

In WMT settings, if you choose "Don't set a preferred domain" then you wouldn't get this.

I can't try that because it says in that place "Restricted to root level domains only".

On a related matter, for "Geographic Target" it says "Your site's domain is currently associated with the target: New Zealand" and I can't change it.

I get the impression that Google doesn't understand .net.nz as a TLD but rather thinks it's a subdomain. Whether this has any bearing on the above issue I don't know.

I don't want separate reports

However you do it, you WILL get separate reports.

For incoming links the report for the non-www will list links that point at the non-www. The report for the www will list links that point at the www.

The crawl report for the non-www will show how many non-www URLs Google asked for per day. The crawl report for the www will show how many www URLs Google asked for per day.

It couldn't work any other way.

What? That is the whole point of canonicalization... to not have to deal with two sets of non-distinct data...

I have never seen this before. Are you seriously saying that everyone has two Google WMT reports for each of their domains...?

Yes. There are two reports.

Canonicalisation simply redirects requests for the non-canonical domain to the canonical domain. That neans that only one version appears in the SERPs.

However, I often look at the non-www report merely to pick up sites that have linked to example.com instead of to www.example.com and then drop them a note to fix their link.

if google can resolve the verification file for both hostnames (a simple test) it is assumed you also have control of the non-www hostname.

The verification atm is the verify-v1 tag in the index file, i.e. the HTML at the root; i.e. the headers above show the 301 working for this.

regardless of the method used, if google sees a common verification for multiple requested hostnames of a root domain it is assumed you have control of those hostnames.
with or without a proper canonical hostname redirect...

I can't try that because it says in that place "Restricted to root level domains only".

example.net.nz is a third level domain, not a root domain.
example.nz is a root domain.

On a related matter, for "Geographic Target" it says "Your site's domain is currently associated with the target: New Zealand" and I can't change it.

ccTLDs such as .nz have an implied geographic target, so google only permits geotargeting for gTLDs such as .com or .net.
Geotargeting - Webmaster Tools Help:
http://www.google.com/support/webmasters/bin/answer.py?hl=en&answer=62399 [google.com]

So I should add the report. I have been ignoring the message all this time.

example.net.nz is a third level domain, not a root domain.
example.nz is a root domain.

I know, but isn't it more important whether it's a TLD or not? You're seriously saying it won't let you set the preference for example.co.uk whereas it will example.com?

The .uk and .nz already have a pre-defined location assumed. You can't change them.

from the GWT geotargeting reference to which i linked above:

Sites with country-coded top-level domains (such as .ie) are already associated with a geographic region, in this case Ireland. In this case, you won't be able to specify a geographic location.

I meant the "preferred domain".

I just ignore those requests, as you have been doing so far - and I see no problems from that. However, I do think that the future belongs to "no www" canonicalization, for a whole host of reasons.

However, I often look at the non-www report merely to pick up sites that have linked to example.com instead of to www.example.com and then drop them a note to fix their link.

You don't have to. I think the very purpose of setting a preferred domain is to tell Google to treat all links to the other canonical form as links to the form you prefer.

But they also expect people to 301 redirect one canonical form of links to the other. I am not sure why they expect people to do both.

I don't think they "expect" people to do both, it's just that not everyone even CAN do the canonical 301, depending on their hosting situation. The preferred domain choice is a "for display" choice, and it still leaves wiggle room for technical errors in Google's ranking calculations. Matt Cutts has often confirmed that the redirect is the best choice.

I do think that the future belongs to "no www" canonicalization, for a whole host of reasons.

Care to expand on that?
I always considered www to be a useful denominator for a website, as opposed to any other service or protocol a domain might be serving.

Essentially, the www doesn't communicate much useful information but it does take up extra space on-screen and in print. The letters represent a hostname not a protocol, and shorter is usually sweeter when it comes to marketing communication, as long as the essential information gets conveyed.

Also, I'd rather see more characters used for displaying the filepath in the search results. In short, "www" is like an old bad habit that we can do without.

I don't set a preferred domain in WMT. That setting works only for Google. I set it for all users and for all searchengines by using a 301 redirect on the site.

If you don't set a redirect, you'll find your link equity being split more and more as people simply cut and paste the URL they see in the browser address bar. If you set a redirect, it doesn't matter what you type in or click on: the displayed URL will be the canonical version every time.

I do think that the future belongs to "no www" canonicalization, for a whole host of reasons.

I think it belongs to the www, merely because of the sheer usefulness of these two searches: [site:www.example.com] and [site:example.com -inurl:www].

You can still brand on "example.com" and when it is typed in, you simply redirect to add the www. Google.com does that. Very many sites do that.

How is that search so useful g1, and why can't it be replaced with the opposite:
site:example.com +inurl:www
For a cannonical example.com that will show links incorrectly pointing at www.example.com, which is the point yes?

The letters represent a hostname not a protocol

I don't think it's really relevent, because the protocol format is too messy and long. Nobody reproduces http:// or ftp:// or whatever, and if they do it's unnecessary with a standard prefix. You just use www.example.com and ftp.example.com and there's no ambiguity. Problem comes from talking about just "example.com" when multiple services are being provided.

The site: search is all about getting a list of indexed URLs. It is NOT about listing links pointing to the site.

If the canonical domain is www.example.com:
[site:www.example.com] shows only the canonical URLs within a site.
[site:example.com -inurl:www] shows only the non-canonical URLs within a site.
This allows you to easily find any URLs which need to be redirected, such as those at example.com and foobar.example.com.

For your alternative where the non-www is the canonical domain (i.e. bare example.com):
[site:example.com] shows both the canonical and non-canonical URLs.
There's no search you can do, to show ONLY the non-canonical URLs.
It is therefore likely that you will miss some of the non-canonical URLs if the list of canonical URLs is very long.

Sure, [site:www.example.com] or [site:example.com +inurl:www] will find www URLs but it won't find foobar.example.com nor any other of the other infinite non-canonical possibilities.

What is needed is a search that includes anything.example.com, while excluding bare example.com URLs. It can't be done.

That's why I prefer www.example.com as the canonical domain. And, as I mentioned earlier, there is no problem whatsoever in branding as example.com while using www.example.com for the site URLs.

See also: [google.com...]

If you are seeing no data for the site www.example.com, it may be because you have added your site using http://example.com. To Google, these are entirely different sites. If you feel like you're missing some data, add both the www and the non-www version of your domain to your Webmaster Tools account. Take a look at the data for both sites.

To Google, these are entirely different sites

God knows why, when a 301 is telling it different?

I also don't understand why I never saw that message for another .com I had - only ever had the www.example.com that I added. Does Google not 'discover' the other domain unless it's led to it somehow?

Anyway, I added the example.com domain, and after about 48 hours there's no data for it.

Yes, they only discover other domains if something somewhere links to them.

There will be very little data in the "other" report. Ideally there will be none. However you should check it to make sure that no-one links to the non-canonical domain.

Your 301 redirect delivers the visitor to the right place if they do, and Google attempts to combine the links for ranking purposes. However, you can improve the situation by actively asking those other sites to correct their links.

The 301 doesn't exactly tell Google that two things "are the same". It says "don't index this one, index the other one".

The "rel=canonical" attribute tells Google that two things should be treated as the same thing, and since they can pull both pages they can verify that, and then act on it or ignore it.

The "rel=canonical" attribute doesn't stop people linking to the wrong version, as the browser URL bar does not change. That's a disadvantage and something that a proper redirect heads off.