Re: UserPrefsDialog results in jsonp call to http://gmodules.com/ig?

[Paul Lindner <li...@inuus.com>]

[responding waaay after the fact...]

The metadata call is one of the few that does support jsonp.   If you've
been following along you'll see that the JSON-RPC endpoint supports it now
and I have a patch review up for REST support.

Next step is to convert the metadata servlet into a proper 'Apps' handler
that can be used with JSON-RPC and/or REST, eliminating one more custom
piece of code/logic.

On Tue, Mar 9, 2010 at 1:28 AM, Tim Wintle <ti...@teamrubber.com>wrote:

> On Mon, 2010-03-08 at 14:25 -0800, Paul Lindner wrote:
> >
> > I'd love to see someone take this patch and make it generic..  If I
> > had some
> > spare cycles I'd do it myself...
>
> Unless I've missed something, it doesn't seem practical to have the
> option of building the user pref UI client side unless the metadata call
> supports JSONP or similar for X-domain metadata requests.
>
>
> I.e. for this patch to work the javascript gadget object must have been
> created with the correct user pref schema - which (as far as I know) can
> either be:
>
>  * Fetched from a container which already knows the user pref schema for
> that gadget (in which case it could have templated the form cheaply
> server-side).
>
>  * Fetched via a POST metadata request. (with X-domain issues
> surrounding this)
>
> We've used both of these techniques here - The server side technique is
> more customizable and IMO better, but it can't be used on static pages,
> and it either leads to a slight extra delay as the metadata request is
> made, or (if cached) to noticable data integrity errors if the gadget
> spec is updated in shindig's cache but not in the container's cache.
>
> The second technique requires a hidden gadget inserted onto the page
> which can make the metadata call (as it's on the same domain), and then
> use gadgets.rpc to send the schema back to the container.
>
> As I say, I may have missed an important call at some point that may
> have made this easier - but it seems allowing cross-domain access to the
> metadata call would be preferable in the second situation.
>
> Tim
>
>
>