[whatwg] Where did the "rev" attribute go?

Ian Hickson ian at hixie.ch
Wed Jul 5 17:49:39 PDT 2006

On Wed, 5 Jul 2006, Charles Iliya Krempeaux wrote:
> It would be a shame to get rid of it now, now that web developers are 
> starting to become "semantically minded".

On the contrary, I would argue that we should get rid of it as fast as 
possible, so that we don't scare away authors who are becoming 
"semantically minded" by making the language more complicated than 
absolutely necessary.

> As developers start building semantics into web technologies, their 
> going find that they need the "rev" attribute.  (Not sure if that would 
> be enough "justification" here to keep it.  But since we already have 
> it, it would be nice to keep it.)

For HTML5 the assumption is that we're removing everything unless we can 
put forward a convincing argument to keep it.

What are the use cases for "rev"? Do they outweigh the author cost?

> Without the "rev", we could still do this with "rel" however, the exact 
> same label could NOT be used.  And then any kind of system to analyse 
> this would have to understand what this labels actually meant and know 
> what the "reverse relation" name was.

It would seem to me like that's the case anyway, for any practical 
application. Other than the software you have mentioned, I have never even 
heard of any generic-purpose parsers for this kind of stuff -- and if such 
a parser was going to be useful, you'd think that in the 16 or so years 
since HTML was invented, someone would have made one and people would be 
using it.

> That is essentially my personal reason for wanting "rev".  So that the 
> exact same label can be used no matter what the direction of the 
> relation.  (That way a generic parser and query engine can be written 
> for this type of stuff, even if the system does NOT understand the 
> meaning of the labels.)

While I understand what you're saying, it seems highly theoretical. Data 
on authoring practice shows that authors simply don't understand rev="": 
the top five <link rev=""> were made, stylesheet, owns, author, and owner. 
rev="made" is the only one of those where rev="" wasn't a typo for rel="".

Usage of <a rev=""> was so low that even the "height" attribute on the 
"space" element is used more often, according to the data I have. And the 
<space> element doesn't even exist! In contrast, the <a rel=""> attribute 
was used more than 60 times more often than the <space height="">. (I use 
<space height=""> as the example here because that's the least-used 
attribute that I actually recorded data for; <a rev=""> is used so rarely 
that it didn't even appear on my top-1000 attributes list. I have data 
regarding <a rev=""> values because I specifically recordeded rel/rev data 
in the study, to determine whether or not we should keep rev="".)

(This data is based on a crawl of approximately one billion documents.)

> I'm using it for trust metrics and reputation.  (But that's just me and 
> the software I write :-)  )

Could you give us some pointers to this software?

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

More information about the whatwg mailing list