[whatwg] WA1: rev attribute

[Ian Hickson]

On Mon, 18 Jul 2005, fantasai wrote:
>
> The 'rev' attribute from prior versions of HTML is missing in WA1, and I 
> think it deserves not to be left out. Most common link types out there 
> are used with 'rel', but some 'rev' values can also be useful.

Actually, research suggests that rev= is basically only used when authors 
make mistakes. The most common rev="" value is "made" (11 million pages 
out of a billion pages), which is redundant with rel=author, and the 
second most common one is "stylesheet" (2 million out of a billion pages), 
which is a mistake. The next most common value is "owns" (110 thousand 
pages out of a billion), which I don't understand, and the fourth most 
common value is "author" (50 thousand out of a billion), which seems to be 
a mistake as well.

This is a very poor track record for a feature.


> Here are some use cases:
>   - rev="footnote" for a link back from the footnote or endnote to
>     the source anchor in the main text

This could be handled by a rel value, e.g. rel=source.


>   - rev="help" for a link to the part of the site that the help
>     text is about

...as could this.


>   - rev="author" on a personal site or resume for links to documents
>     s/he has written

Why need a rel type at all for this?


> See also http://www.eastgate.com/HypertextNow/archives/Trigg.html
> for a direction link types could go in which 'rev' would be useful.
> Many of the link types suggested there would be easier to use with
> rev for the reverse link than with a separate keyword that means
> the inverse relationship.
> Example:
>   rev="refutation" to link to the article one is refuting

I think we should work the other way around. We should start with a 
problem, and work out the solution, not the other way around.


On Mon, 18 Jul 2005, Matthew Raymond wrote:
> >
> >    - rev="footnote" for a link back from the footnote or endnote to
> >      the source anchor in the main text
> >    - rev="help" for a link to the part of the site that the help
> >      text is about
> 
> This is largely useless, as you are unlikely to start at a help/footnote 
> document and go to the document for which the help document was written. 
> The most common situation is that you clicked the help/footnote like 
> from the parent document, and therefore the relationship is already 
> established from the parent document.

I agree.


> >    - rev="author" on a personal site or resume for links to documents
> >      s/he has written
> 
> Here, you're using |rev| to replace missing metadata in the target 
> document. What happens when <meta name="Author"> is defined in the 
> target documents? Does |rev| override? What would a UA do with the 
> information anyway? If there's a link, wouldn't there be text stating 
> that the creator of the personal site created the document the link is 
> to?

Indeed.


> > See also http://www.eastgate.com/HypertextNow/archives/Trigg.html
> > for a direction link types could go in which 'rev' would be useful.
> 
> Well, I scanned over it, and I noticed one good point. People often 
> don't bother putting in relationship types for links. Therefore, |rev| 
> could establish what the relationship is when you reach the target 
> document. The problem is that the argument is mostly self-defeating. If 
> people fail to use |rel|, how is a reverse version of that same 
> attribute going to be used with any frequency.

Evidence suggests it is not.


> At least with |rel|, you could harvest hyperlinks and put them into a 
> link toolbar. With |rev|, you're describing the relationship type of the 
> current document. Therefore, I really don't see what user agents are 
> supposed to do with |rev| and how they can create a useful interface 
> that can exploit this attribute.

Indeed.


On Mon, 18 Jul 2005 sjoerd at w3future.com wrote:
> 
> The user interface of rel and rev can be exactly the same, only rev 
> under the heading of "reverse".

In practice, it seems authors and users alike don't really care for this.


> AFAIK there is no difference between
> 
>   <a href="1.html" rel="prev">
> 
> and
> 
>   <a href="1.html" rev="next">

Indeed. So why have rev=next?


On Mon, 18 Jul 2005, Matthew Raymond wrote:
> 
> So, functionally, you're just breaking a link toolbar into two 
> categories: "forward" and "reverse". What's the use case for this? 
> Surely a "Previous" button in your links toolbar is better than 
> "Reverse->Next" from a UI perspective. Or are you suggesting that the UA 
> should determine the reverse of the relationship and present a button 
> for it? That's really bad for things that don't necessarily have 
> inverses:
> 
>   |rev="top"| -> "bottom"?
>   |rev="first"| -> "last"?
>   |rev="top"| -> "bottom"?
>   |rev="ToC"| -> ??????????

Indeed.


> Also note that "refutation" is a bad example, as it would only ever be 
> used in |rev|. Does anyone ever link to a refutation of their article 
> from the article itself???

Good point!


> So what we're seeing is that |rev| encourages us to define relationship 
> types specifically for |rev| that are useless for |rel|.
>
>    Another thing is that |rev| is largely self-serving:
> 
> | <a href="http://whatwg.org" rev="supreme-master-guru">
> 
> By it's nature, |rev| defines how the universe relates to you. Thus, how 
> can you help but put yourself at the center of the universe?

A point worth considering indeed.


On Tue, 19 Jul 2005, fantasai wrote:
> 
> Or maybe I just scrolled to the bottom after reading the whole text 
> straight through and want to jump to the context of the footnote I'm now 
> reading. (The footnote and its context could be in the same document, 
> too, y'know.)

It doesn't seem like a rel="" is needed for this. The link itself is clear 
enough (see, e.g., wikipedia).

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'