[uf-discuss] mixing vocabularies

[Ben Ward]

Hi Peter,

On 24 Jun 2009, at 18:55, Peter Mika wrote:

> Look at for example at [1]. This page contains the following markup:
>
> <table class="infobox infobox vcard vevent" cellspacing="5"  
> style="width: 22em; text-align: left; font-size: 88%; line-height:  
> 1.5em; font-size:90%; text-align:left;">
> <tr>
> <td colspan="2" class="fn summary" style="text-align:center; font- 
> size: 125%; font-weight: bold; font-size:110%;  
> background:khaki;">Kevin Bacon</td>
> </tr>
>
> If I look at it strictly, I have a vcard and an event object which  
> both have the name "Kevin Bacon". However, what the author intended  
> is probably a person object, with some terms borrowed from vevent  
> (not sure which ones).

So, in this case the vevent in that page — http://www.answers.com/topic/kevin-bacon 
  — is invalid — certainly incomplete. That structure doesn't contain  
any dates at all.

What I posit has happened is that at one point, answers.com marked up  
the ‘Years Active’ part of that info box with dtstart and dtend.

They're not marking up one object with a combined vocabulary, they're  
marking up two objects: One card (for Kevin Bacon) and one event  
(Kevin Bacon's Career).

I think they backed out dates at some point, but have left the root  
class and summary class in place. With the dates in place, two  
distinct but valid objects would be parsed out.

Answers.com could instruct someone on how to parse the two  
microformats in combination for additional context, but the structures  
standalone too.

> So what do you guys think about this? Note that on our side this  
> introduces the secondary problem that we now have to figure which  
> object is the main topic of the page (it's very clear for the human!)

Figuring out ‘the microformat for the page’ is not a consequence of  
‘mixing vocabularies’ in this context; that is, overlapping or  
integrated structures. It's a problem presented when you have multiple  
objects (of any structured data origin) anywhere in the same page.  
That's a really interesting problem in itself, but not directly  
correlated with this one.

I'll start up a brainstorming page for that though; we talked about it  
with the other SearchMonkey guys at the µf dinner a few weeks ago.

Cheers,

Ben