[microformats-discuss] XHTML tables as CSV-like "records"?

[Ryan King]

On Oct 2, 2005, at 9:17 PM, Dr. Ernie Prabhakar wrote:

> Hi Tantek et al,
>
> On Oct 2, 2005, at 1:30 PM, Tantek Çelik wrote:
>
>> Right.  Before jumping to a microformat, we should always first  
>> ask, can I
>> "just" use XHTML elements to do this?  And if not, can I create a  
>> simple
>> XHTML compound to do this?
>
> I'm not sure I understand the distinction being made (I'm still  
> trying to figure out the terminology, I guess).   I appreciated all  
> the comments, but the most relevant for me was Kevin's:

In short: XHTML compounds are combinations of two elements to create  
a new construct with specific semantics. Nothing is added, only  
recombined. For example:

<pre><code>
...
</code></pre>

Combines the semantics of <pre> with the semantics of <code> to  
create a compound construct.

> On Oct 2, 2005, at 12:00 PM, Kevin Marks wrote:
>
>> This is something we discussed last year in 'Can your website be  
>> your API?',
>>
>> http://tantek.com/presentations/20040928sdforumws/semantic-xhtml.html
>>
>> I think it is worth writing out a set of rules, as we did for  
>> XOXO. In particular, XOXO constrains each <dt> to one <dd> for  
>> simplicitly of dictionary mapping, and constraining to have a  
>> uniform grid of rows and columns makes sense.
>>
>> You can represent n=dimensional arrays in XOXO as nested lists,  
>> but a 2d table is a very useful special case.
>>
>
> That is precisely my point.  I want to know:
>
> a) How -best- to use *existing* XHTML table markup to indicate that  
> idea of a nested list (or list of dictionaries)

Which is a very good question.

> As (I think) Ryan said:

I think it was Brian.

> On Oct 2, 2005, at 11:33 AM, Ryan King wrote:
>
>> the talk also discusses some of
>> the lesser know attributes already build into tables, such as AXIS,
>> HEADER, COLGROUP, ROWGROUP, ID, etc[1]. Even some of the regular
>> elements, TBODY, TH, etc. These already have some semantic  
>> meaning, so
>> a new microformat might not be needed for this application
>
> Yes, I'm pretty sure we don't need new markup *inside* the table --  
> the existing semantics should work fine.  However, I *do* think we  
> need conventions about *which* semantic elements to use, and how to  
> interpret them.  For example, should "scope" always be explicitly  
> specified?   Should we expect *all* conforming tables to have a  
> header row (using <th>)?  Or, how do the semantics change if a  
> given element is or is not present (e.g., does it become a 'tuple'  
> instead of a 'dictionary' if there's no headers?).

These are very good questions, for which I don't think there are any  
thorough, established answers (not that people haven't addressed them  
in many contexts, but we don't have a way for doing it in a lossless- 
xhtml way).

> Which brings me to my second point
>
> b) What differentiates a microformat from a design pattern from an  
> HTML compound?
>
> Is it just the presence of a class name?   In this case -- as at  
> least Kevin seems to agree, even if nobody else does -- doesn't it  
> make as much sense to have a unique classname for structured  
> tables, analogous to "xoxo" for lists? To show that the conventions  
> *are* being followed?

I would break them up thusly:

microformat - an extension to xhtml, usually by adding a class, rel  
or rev attributes. Some specify which elements must be used (eg,  
xoxo), others don't

design pattern - good xhtml practice which emerge over time, which  
are sometimes described formally

html compound - see above

> Or is there some subtlety here I'm missing?
>
> Conversely, could we conceivable add 'table' as an element of XOXO  
> (explicitly defined as a 'compact' representation of multiple lists)?

Actually, it seems we need something like xoxo for tables:

xoxo is to outlines as ? is to tables.

> I'm not trying to 'hurry' into anything -- I'm genuinely confused,  
> and open to any and all suggestions. :-)

Your questions are good and well thought out. I hope my answers have  
been helpful.

-ryan
--
Ryan King
ryan at technorati.com