Re: [link draft] Changing the model for links

On 09/04/2009, at 4:37 AM, Ian Hickson wrote:
> [snip]
> Similarly, the position of <link> elements relative to other elements
> (e.g. <style>, <script>) and HTTP Link: headers is relevant to  
> behaviour.

Making this explicit shouldn't present a problem. What's the preferred  
precedence in HTML5?

> On Wed, 8 Apr 2009, Phil Archer wrote:
>> For me, the biggest hole in Link in this respect is the lack of  
>> support
>> for media types. Not being able to write
>> Link <styles1.css> media="handheld" ...
>> Link <styles2.css> media="screen" ...
>> is a real downer but I know this has been raised before.
> Yeah, it seems like we would want provide attributes like media=""...
> HTML5 introduces sizes="" for one rel-type, if HTML is to use the same
> model so that you can set favicons from Link headers also, then we'd  
> have
> to include this too.

Yet again, the draft allows content-specific extensions like media.

Is there an argument that 'media' is generic to all links, and is  
there a content-neutral way to describe it? If so, we can get it in  

> On Wed, 8 Apr 2009, Mark Nottingham wrote:
>> What's the status of "up up", etc. in HTML5 these days?
> Tentative. (The section is marked "last call for comments".)

Ack. I'll make a comment.

> I think it would be very unfortunate to lost this feature just  
> because the
> underlying model can't describe it, though.

I think we're hearing more than that -- that it's bad practice to make  
relations behave like this, because it's tying them to a very specific  
syntax, and making them operate in a way that's very specific to the  
relation type.

>> Also, has testing been done around UAs with regard to serialisation  
>> in
>> HTML? I.e., which of these qualify as an alternate stylesheet  
>> according
>> to implementations?
>> <link rel="alternate stylesheet" href="a"/>
>> <link rel="stylesheet alternate" href="a"/>
>> <link rel="stylesheet" href="a"/><link rel="alternate" href="a"/>
> The first two. I don't know of any software that does special things  
> with
> rel="" for serialisation, so I haven't tested anything on that side.

Interesting; < 
 > doesn't sanction the second form, so browsers are already taking  
license with the (very weakly implied) model in HTML4.

>> Again, if "alternate stylesheet" is grandfathered in and "up up" is
>> supported by saying that all forms that result in two links to the  
>> same
>> target with 'up' as the relation, it's a lot easier to get there.
> I don't think that would really work, I think it would be quite  
> plausible
> for example for pages to have breadcrumb links at the top and bottom  
> of a
> page.

Sure, but what's the actual use case, beyond being able to mark them up?

> Also, note that as defined in HTML5 the breadcrumb mechanism is  
> scoped to
> the current paragraph. This is something else that might be hard to
> express in a common data model.

No, it works just fine; you just need to define the context of the  
link as the paragraph.

> On Wed, 8 Apr 2009, Mark Nottingham wrote:
>> So, I'm very interested in whether 'up up' is a good idea, whose  
>> idea it
>> is, who wants to implement it, and who already has.
>> More to the point, does anybody really need 'up up up up up up up',
>> which is what this implies?
> Link types in general aren't that widely used, sadly. The idea with  
> the
> new rel-up/rel-index feature in HTML5 is to enable breadcrumbs to be
> expressed; quite a lot of people have asked for ways to do this. I  
> don't
> know if what we have now is the best way to do it.

OK. If you can show us why they want it, or put us in touch with them,  
it would be most helpful.


Mark Nottingham

Received on Wednesday, 8 April 2009 23:40:43 UTC