Add link rel="canonical"

[domenic]

I was planning to use this today, went to the spec to look up the syntax and any usage notes, and found out that it's not there. It seems like something so useful and widespread should be in the list of link rels in the spec. For example we could say that it's only useful on link, not a/area.

It's already in http://microformats.org/wiki/existing-rel-values#HTML5_link_type_extensions which states it can be used on a/area, hmm. Maybe we should just leave it there? But why is canonical in the wiki and author in the spec?

See also https://en.wikipedia.org/wiki/Canonical_link_element

/cc @sideshowbarker

[annevk]

Maybe @tantek has thoughts on this?

[tantek]

We have a stub of a spec: http://microformats.org/wiki/rel-canonical which technically anyone can login and edit (all contributions to microformats wiki are covered by CC0 + OWFa, similar to how contributions to Wikipedia are covered by their contribution license).

Update: IIRC @kevinmarks (who worked at Google when rel=canonical was first becoming a thing) has often had a lot to say about rel=canonical, when to use it instead of new rel values etc.

Kevin, want to take a crack at expanding http://microformats.org/wiki/rel-canonical with your thoughts on meaning, usage, when authors should use it, what should browsers, search engines, and other consuming applications do with it?

[tantek]

Also re: "something so useful and widespread should be in the list of link rels in the spec" what I remember is that we gave up trying to maintain an explicit in-spec list beyond what went into HTML5 because it became far more "formal" work than necessary. Hence the easy to edit registry instead. If there's a good reason to reconsider growing an explicit list in the HTML spec, I'm not against it, I'd just be curious about what reasoning has changed since we gave up doing that.

[kevinmarks]

I've updated http://microformats.org/wiki/rel-canonical with a more thorough description and discussion - further edits welcomed.

[sideshowbarker]

I added a use counter to the W3C HTML checker to collect data on usage of rel keywords. That has so far recorded data on 1,268,156 unique documents checked. Based on that data, the table below ranks the keywords based on how often they were found among those 1,268,156 documents.

As the table shows, canonical was found in almost 32% of the documents checked, and that makes canonical the third-most-commonly-used rel keyword—just after stylesheet and icon.

---

Rank	Keyword	Number	Percent
1	stylesheet	1060847	83.65%
2	icon	712571	56.19%
3	canonical	402446	31.73%
4	alternate	377532	29.77%
5	nofollow	296259	23.36%
6	dns-prefetch	162899	12.85%
7	pingback	139291	10.98%
8	next	83834	6.61%
9	search	78826	6.21%
10	bookmark	66861	5.27%
11	tag	66103	5.21%
12	prev	58764	4.63%
13	author	42907	3.38%
14	external	20615	1.63%
15	preconnect	17217	1.35%
16	noopener	10560	0.83%
17	preload	8370	0.66%
18	license	7220	0.57%
19	noreferrer	7078	0.56%
20	prefetch	4869	0.04%
21	help	2966	0.02%
22	prerender	760	0.01%
23	serviceworker	0	0.00%