Re: Producing Atom

Ian Hickson wrote:
> On Sun, 4 Oct 2009, Sam Ruby wrote:
>>> (Incidentaly, a side-effect of this is that the HTML-to-Atom 
>>> conversion algorithm can no longer output valid Atom. It used to rely 
>>> on the vCard vocabulary to get the value of <author>, but this is no 
>>> longer possible since there's no reference to the vocabulary specs.)
>> re: "no longer possible"... why not?
> Atom requires an <author> element, but the only mechanism in HTML to 
> specify an author's name as part of contact information would be to use 
> Microdata (a vCard in an <address>). Since HTML5 hosts the Atom conversion 
> algorithm but vCard is no longer in HTML5, I had to remove the part of the 
> Atom conversion algorithm that referred to vCards, thus preventing the 
> required <author> element(s) from being generated.

Microdata is certainly a mechanism for specifying Author information.

>> In particular, what would be the harm in allowing conforming 
>> implementations to add additional information as long as they continue 
>> to provide all of the information required by the algorithm?
> People might come to rely on the information added by particular vendors, 
> at which point we effectively have changed the de-facto algorithm and 
> would have to update the de-jure algorithm also, regardless of which specs 
> that meant we had to reference.
> But that's not particularly a big problem, at least not in my opinion. 


> It's also not really relevant: the point isn't that it's not possible to 
> have an HTML-to-Atom conversion that generates valid Atom (it obviously 
> is, since HTML5 used to have such an algorithm); the point is that with 
> the restriction that we can't reference the vCard algorithm from HTML5, 
> the algorithm had to be changed in such a way that the output is no longer 
> conforming by default.
>> Furthermore, why couldn't additional specifications define additional 
>> information to be placed into the feed?
> They could, but that wouldn't change whether HTML5's algorithm alone 
> produced conforming Atom.

What does an HTML5 to Atom algorithm need to be in the HTML5 spec?

If it were in a separate spec, it could normatively reference the vCard 

- Sam Ruby

Received on Sunday, 4 October 2009 21:18:37 UTC