[dev] Horde_Feed api evolution

[Chuck Hagenbuch]

Hi folks-

I'd like feedback on some plans to change the Horde_Feed API a little  
to make generating feed content (as well as reading it) easier, and to  
generally clean up/improve the API.

First proposal: remove the code in Horde_Feed_Abstract to load feeds  
via HTTP. We have Horde_Feed::readUri() and the other  
Horde_Feed::read* methods for that.

Second, I'd like to switch the order of $uri and $element in the  
Horde_Feed_Abstract constructor, so that $element comes first. With  
$uri not used for reading anymore, it really is optional, and $element  
can be used to pass in initial data when instantiating a feed  
directly. This also goes into my last proposal:

Allow $element to be an array of simple children to set on the feed  
(or entry - this functionality would probably go in to  
Horde_Xml_Element). I'm proposing that $element could be a simple  
name/value array, with a syntax of elementname#attributename. Easier  
to show it:

$atom = new Horde_Feed_Atom(array('link#href=foo',
                                   '#xmlns=rss'));

This would result in the following XML:

<feed xmlns="rss">
  <link href="foo" />
</feed>

In other words, # is translated into an attribute name, either on a  
child or on the feed itself.

Thoughts? Any other ideas for changes/improvements to the feed api?

-chuck