Re: data model and example

On Sep 30, at 12:35 AM, Rob Manson wrote:

> Hey Raj,
> 
>> No. I wanted to specify the author, but I'm having trouble figuring
>> out how to represent it. See:
>> http://rajsingh.org/poiwg/poi_logan_2.xml
> 
> Yeah I think that's better.  I'd assume if <value> is not defined then
> the full content is assumed to be the payload?

I suppose that's a possibility. It would look cleaner, but as a developer, I'd rather just always know to parse the <value> element as text, rather than have a conditional check to see if there are elements within <description>, and if not then treat <description> as text.

>>> Is the <point><Point>...</Point></point> double nesting needed?
>> 
>> Yes. Everything in <Point> is straight from GML 3.3. And the
>> double-nesting will be useful. For one, it will avoid the problem with
>> <description> above, where you don't want to mix CDATA with an
>> element. I.e. it's easy to slip an <author> element in as a child of
>> <point> without messing with the location specification in <Point>.
> 
> Fair enough.  I'd also like to be able to support the simpler geo:uri
> too as a lot of developers would be happy with just that too.

At the face-to-face we decided to put a stake in the ground here and have a single way to represent a point. There are too many "pet" point formats around. It would be better in the long run for the world to focus on a single one for interoperability. And what better place to have the canonical one than the W3C POI spec? And for practical reasons, if you're generating POIs, writing <Point><posList> instead of geo:uri is the least of your problems! And as a consumer of POI data (at least the XML format), it's nice to always know to find the coordinates in the <posList> element.

Received on Friday, 30 September 2011 18:23:04 UTC