[microformats-discuss] URIs please!

[Ryan King]

On Jul 14, 2005, at 8:14 AM, Bud Gibson wrote:
> On Jul 14, 2005, at 9:33, Peter Janes wrote:
>> I've had a bit of previous offline discussion with Tantek, and  
>> it's been my impression that he's in favour of requiring linked  
>> profiles (since that was exactly what we discussed).  I think the  
>> rejection of "Just because a profile value mentioned in a  
>> microformat's linked XMDP also appears in the document does not  
>> mean that that microformat is in use." is more one of wording/ 
>> interpretation, since it could be taken to mean:
>>
>> - "just because I've listed the XFN profile doesn't mean that I  
>> use any of the values defined in it"
>> - "just because I've listed the XFN profile doesn't mean that  
>> rel='met' means what it says in XFN" or
>>
>> I think the purpose of the statement was to express the former  
>> argument, and the objection is to interpreting it as the latter.
>
> I agree that we probably need to clear up interpretations which is  
> why I am continuing to press this issue.  Having reread the spec a  
> couple of times, here is my best interpretation of what linking to  
> an XMDP means:
>
> 1.  The presence of an XMDP profile only means that a particular  
> microformat may be present in the document (Danny Ayers' point).

You're right, it doesn't necessarily mean that the microformat  
appears in the document. But it it does show up, then the semantics  
of the profile should be applied to it.

> 2.  The presence of an XMDP cannot practically be assumed to mean  
> that any occurrence of an attribute value defined in the XMDP is to  
> be interpreted according to the XMDP's definition.

Right, but we're not looking for individual attribute values, we're  
looking for attribute values in context. So, given that an hreview  
profile has been referenced, this would not be interpreted as an  
hreview description:

<div class="description>...</div>

but this would

<div class="hreview><span class="description"></span></div>

>   There are two simple cases that suggest this point:
> a.  Two XMDPs are referenced, each one referring to the attribute  
> value.  Which takes precedence?

I believe this has already been covered in the XMDP spec– the  
profiles are applied in order, so the first one takes precedence.  
This is assuming that there really is an ambiguity that is not  
cleared up by the context in quick a microformat is used.

>   I can come up with some rules of thumb, some of which I have  
> heard at conferences, but where are those explicitly stated?

Perhaps nowhere.

> My point here simply is that there is a lot of unwritten lore that  
> will leave newcomers guessing.

Right, and feel free to bring these issues up.

> b.  The fact that XMDPs appear at least implicitly to be oriented  
> toward defining a specific context that may occur in part of a  
> document.  While in discussions concerning the development of  
> xFolk, I initially attempted to create unique attribute value names  
> and was told that this was "unnecessary syntactic vinegar", the  
> attribute value names would be interpreted in context.

Right.

> 3.  The XMDP does not give explicit syntactic significance to the  
> attribute value that indicates when it is to be interpreted as  
> applying.  Therefore, without reading the XMDP, knowing when it is  
> actually in effect is impossible.  This is not just "humans first",  
> it is "humans only".

Right. XMDP's value is almost exclusively for humans. But I think  
that's much more important at this point.

> Given all these points, XMDPs need refinement before they can be  
> used for automated microformat discovery.

I think you're missing the point of the XMDP. The XMDP is largely for  
documentation. Its the profile url that is used to identify when that  
profile's semantics should be applied to a document.

> At this stage, the only available discovery method is to hand-code  
> the attributes that indicate particular microformat contexts into  
> your parser and assume some rules of thumb.  This method is still  
> viable at the current scale.
>
> Boy, it seems to make autodiscovery harder than it needs to be,

I still have no idea what people mean by 'autodiscovery.'

> particularly as microfomats proliferate.  The simplest solution  
> would seem to be to syntactically mark the microformat's enclosing  
> element in the XMDP so that it can be extracted during the parsing  
> process.

During which parsing process? Parsing the XMDP?

-ryan