Since you want to get away from the html 5 arguments, I’ll assume we can agree for arguments sake that html 5 was incorrect in removing @rev, and will shortly reintroduce it.

So your central argument is that rel=short* is better than rev=canonical to specify short url alternatives, correct?

I would argue that a rel=short* is by definition canonical, so if we can imagine @rev is alive and well, you would use rev=canonical for every rel=short* because rel=short* is a subset of rev=canonical.

In a world of documents where every rel=short* has a rev=canonical, rel=short* still has a use I would argue, which is to specify the publishers preferred short url, but it’s a marginal benefit really, because it is not hard to deduce which of the 1+ rev=canonical’s is the shortest, or short enough to satisfy the user.

Some argue that if rev=canonical is used at all it should be used for every rev=canonical, a possibly infinite number. I disagree, and say the publisher should simply use rev=canonical for any rev=canonical link that may be of interest to the user. This will enable many more discussions between a user and a document to occur then the one we are all dealing with this month, which is providing a short url.

In response to your dialog above, I think you went astray right from the start, try this use case:

User: “Hey document, your canonical url is f***ing huge, does you canonical url represent any other urls I can scan for a shorter version I can use?”

Doc: “Yeah sure here are some.”

User: “Nice I picked one short enough, dope!”

[If short url is from another domain]
User [thinks]: “Wait this url is from another domain.. is it safe?”

User [thinks]: “I better verify by checking it’s http headers for a rel=canonical or a 301 redirect to a page the canonical represents, if there is not one there then I can also scan the html if I’m a keener.”

User [thinks]: “done, off to twitter.”

User [thinks]: “those old days of using tinyurl were so lame.”

Well actually a user is not going to check the http headers, but they could trust the rev=canonical, or use a tool which will do all of the above for the user. Like say a bookmarklet or a ubiquity command, or a firefox extension, etc..

“Hi there, HTML page. I found you with this URL I’ve got handy. I like you, but this particular URL is a bit wide. You got any others I can use to find you?”

“Sure, I’ve got at least one listed as rev=canonical in my head. There might be more.”

“Oh, hey, I found the list and at least one of them is shorter than what I had before. Oh yeah, and we don’t like the ev.il URL shortener as a policy on our end, so we’ll use that good.ie link instead.”

The crux of my argument is choice, both for the publisher and for the link carrier. Granted the ev.il vs good.ie service choice is a tad contrived – but as you can see, URL length preferences and acceptable service policies are supported by rev=”canonical”

But – and please don’t take this personally – I don’t agree that any of this matters in the first place. It’s a real storm in a teacup.

The only reason there’s a debate about how to represent “short” URLs is because Twitter is fairly popular among developers right now. Twitter is one small (and frankly quite annoying) part of the Web ecosystem, and people are proposing changing HTML to accommodate it, in ways that will be completely ignored by every user and user agent out there, and splitting hairs about how best to do so? Really? Does this deserve the amount of noise it has generated?