[whatwg] Absent rev?

[Ian Hickson]

On Tue, 18 Nov 2008, Martin McEvoy wrote:
> >
> > We did some studies and found that the attribute was almost never 
> > used, and most of the time, when it was used, it was a typo where 
> > someone meant to write rel="" but wrote rev="". To be precise, the 
> > most commonly used value was rev="made", which is equivalent to 
> > rel="author" and thus was not a convincing use case.
>
> !! rel-author doesn't mean the same as rev-made eg:
> 
> "I have just finished this new <a rel="author" 
> href="http://coolsite.co.uk/">Cool website</a> check it out""
>
> that would mean <http://coolsite.co.uk/> is the author of the referring 
> page which is nonsense. rev="author" is clearly better semantics in the 
> above case?

As defined:

   http://www.whatwg.org/specs/web-apps/current-work/#link-type-author

rel="author" means that the referenced document provides information about 
the author of the section containing the link.


> > The second most common value was rev="stylesheet", which is 
> > meaningless and obviously meant to be rel="stylesheet".
>
> And that was the basis of the whatwg decision to drop rev?

Yes.


> (I am not criticizing just trying to understand it) surely all it needed 
> was to define some rev values (the same as rel) and people will start 
> using rev correctly?

That's backwards -- looking for a problem to fit the solution, not looking 
for a solution to fit the problem.


> > Anything that could be done with rev="" can be done with rel="" with 
> > an opposite keyword, so this omission should be easy to handle.
> 
> There are some cases where that is just not possible.

It's always possible, since at the extreme you can always just prefix the 
keyword with "rev-" and define it as being the oppoite link relationship.


On Tue, 18 Nov 2008, Dan Brickley wrote:
> > 
> > In which cases doesn't it?  If A is the author of B then B was made by 
> > A, surely?
> 
> Then B contributed to the creation of A, yes. Perhaps not on their own.
> 
> But we need it in the other direction too: can we conclude from { A made 
> B } that { B author A } ?
>
> Not if B isn't textual. Authorship is about writing, but there are many 
> other avenues for human creativity (some of which result in things with 
> URLs, eg. software, images, sounds).
> 
> So there are two complications here, and these are very real world 
> issues, chewing up countless hours in projects like Dublin Core.
> 
> First is "a" versus "the". Nothing warrants reading "the" into 
> rel=author. There might be other authors, listed or not listed in their 
> own hyperlink. Or the page pointed to might be a collectively maintained 
> page or group homepage etc. Or a mailto: for a mailing list.
> 
> Second is non-textual creations. The early Dublin Core specs had a 
> "dc:author" property. This was changed back in 1996 or so to be 
> dc:creator, since this better includes visual works, museum artifacts 
> and so forth, ie. things that can be made, but which are not 
> (postmodernism aside) conventionally considered texts. Authorship is a 
> notion that doesn't make much sense in a non-textual context.
> 
> My point in previous mail about shifting work from HTML5 to elsewhere, 
> is that this kind of distinction is subtle for many seemingly obvious 
> pairs of relationship-type names, and that rev= is at least precise in 
> its meaning.

With all due respect, the problem you describe above is not a problem 
faced by many authors, certainly not enough authors to justify keeping an 
attribute used as rarely as, and as incorrectlly as, rev="".


On Tue, 18 Nov 2008, Dan Brickley wrote:
> 
> Without rev, content creators (in every language) will need to go 
> through this dance, hunting through dictionaries and debating 
> subtleties, to make sure that they've identified a suitable pair of 
> words such that { X word1 Y } is true if and only if { Y word1 X }. 
> Which is why I see this in terms of division of labour. Cleaning it out 
> of HTML5 makes work elsewhere...

Just use a convention, e.g. rel="author" and rel="author-rev", if you 
really need this. (They are just opaque strings, after all.) In practice, 
there simply aren't enough people trying to describe reverse relationships 
to make it worth putting the work in HTML.


On Tue, 18 Nov 2008, Martin McEvoy wrote:
>
> Its not explicit enough, there are times when there is a need to express 
> explicit relationships to things, a uniqueness that only you can relate 
> to, rev= is an explicit one way relationship from A to B
> 
> From the "real world" found here:
> http://nfegen.wordpress.com/2008/03/28/micrordformats/
> 
> <p>I read an interesting post recently, <a 
> href="http://internet-apps.blogspot.com/2008/03/so-how-about-using-rdfa-in-microformats.html" 
> title="Link to Mark Birbeck blog post">‘So how about using RDFa in 
> Microformats?’</a>....</p>
> 
> An explicit one way relationship I might like to add to the hyperlink 
> above may be rev="reply"
> 
> <a rev="reply" 
> href="http://internet-apps.blogspot.com/2008/03/so-how-about-using-rdfa-in-microformats.html" 
> title="Link to Mark Birbeck blog post">‘So how about using RDFa in 
> Microformats?’</a>

Just use rel="reply-to".


> That's just a matter of educating people not saying lets take rev away 
> because you don't know how to use it?

"Educating people" hasn't worked well on the Web.


> > What semantics do you think authors who wrote rev=stylesheet were 
> > meaning to convey?  Presumably not that the webpage containing it is 
> > the style-sheet for the CSS file that it linked to -- so it's 
> > definitely a mistake by the author.
>
> It was of course but how many authors make that mistake now?

It's the second most common value after rev="made". Other values were so 
rarely used they're not really even statistically significant.


On Tue, 18 Nov 2008, Martin McEvoy wrote:
>
> That was a good example of how Murky @rel is compared to @rev
> 
> <a rel="in-reply-to"
> href="http://internet-apps.blogspot.com/2008/03/so-how-about-using-rdfa-in-microformats.html"
> title="Link to Mark Birbeck blog post">‘So how about using RDFa in
> Microformats?’</a>
> 
> would be 
> <http://internet-apps.blogspot.com/2008/03/so-how-about-using-rdfa-in-microformats.html> 
> is in reply to the referencing document surely?

It means whatever you define it to mean. You could use the word "banana", 
or a URL like "http://rel.example.org/topic", or a Korean word, it doesn't 
matter so long as you define it as having the meaning you want.


On Tue, 18 Nov 2008, Martin McEvoy wrote:
> 
> No Not that bad rel=muse is near the mark, but the author of the page I 
> am referencing may not give me inspiration, I just want to reply to 
> someone, it may be rhetorical, or insulting?

Why would you need to mark it with a rel="" value at all? What benefit 
does it give you?


> by abandoning @rev you are denying the author the ability to express 
> inverse relationships, the ability to say that I have some explicit 
> relationship to a thing

This is not true. Any relationship B->A can be described as a relationship 
A->B, just define a new term and say it's the opposite of the previous 
term.

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