[whatwg] link-types

Anne van Kesteren fora at annevankesteren.nl
Sun Sep 12 07:37:34 PDT 2004

> I am one of the developer of the "rellinks" plugin of Konqueror (KDE browser). 
> This plugin add a toolbar to manage these links : 
> http://shift.freezope.org/konq_rellinks/

Nice, although I think that every browser should have native support for 
this, like I mentioned in my original post.

> I already make search about the different possible values used actually in the 
> link tags. I also found the similar ones. Here is the result of my search : 
> http://shift.freezope.org/konq_rellinks/development_html  (The "supported" 
> column is just for the status in the plugin).

Nice, that looks like a better survey and mentions a lot more possible 
link-type values. I don't agree with every descision you have made 
(treating 'child' similar to 'next'), but I don't think this is the 
appropriate time to go into discussion about this.

> For links about "link" tag and support of them in tools see 
> http://shift.freezope.org/konq_rellinks/documentation_html . If you have 
> other ones (tools or famous websites) it'll be great if you send them to 
> me :)

I think you have found enough resource already. I think that a sensible 
subset of that should be normatively defined and added to the HTML 5.0 
specification. (I don't really care if there are duplicates like 'start' 
and 'top' ('start' is used by search engines, Mozilla supports 'top') 
that are treated the same.)

> I totally agree that "link" tag need specification. The important things to 
> define are:
> - Which relations are possible ?

The WG should make sure that these relations make sense to be included 
in HTML 5.0. They must be generic, just like the markup languages. 
Specific link-types like the ones defined by XFN should be extensions.

> - Which relations can be multiple ?

A lot. If you are in http://example.org/foo, 'up' and 'top' are the 
same. So you could put them in the same REL attribute.

> - A way to write non-standards relations (like -mozilla-opacity in CSS by 
> prefixing with "-" ?) before their inclusions in the norms.

I believe the proposal mentioned in my previous mail used something like 
'xfn.friend' for extensions. I believe there are ways to define a 
default profile as well.

Note: Mozilla has a '-moz-opacity' property, but now also has a 
'opacity' property.

> - The order of similar relations (example : 2 "top" link. First top is the 
> parent or grand-parent of the page ?).

'top' is the root of the website, similar to 'home' and 'start'. I agree 
that such things should be defined.

> -....
> This specification is important because now all the important browsers support 
> this (except IE)

Actually, while IE doesn't have native support, there is a plugin[1] for 

> so if the specification doesn't appear rapidly then 
> developpers will used their own relation's names that will work with one 
> browser but not the others. Specification can prevent this.

Indeed. The specification should probably define examples as well and 
define in a better way how the REV attribute can be used. REL is 
completely clear for most people, but I never say somebody using REV. 
Actually, I have never seen a good implementation of that attribute.


  Anne van Kesteren

More information about the whatwg mailing list