I think perhaps the underlying assumptions may vary according to the
type of UA?
However, I think even on a single-user O/S (e.g. mobile) or in a
sandboxed user context you would still want to maintain storage of
preferences on a per-instance basis.
For example, if you had more than one instance of a single widget,
you'd most likely want this as you needed to have different
configurations for each (e.g. two single-feed RSS widgets, each one
with a different feed). If widgets shared a single preference state by
type then this wouldn't work.
S
On 25 Apr 2009, at 15:37, Thomas Roessler wrote:
It's probably worthwhile to be more explicit about the requirements
here: In Guido's and my discussion, we assumed a requirement to
have persistent storage that might be available to *all* instances
of a widget. That's different from per-instance storage which could
indeed be solved easily within the currently proposed framework.
I'm not sure whether the current requirements document actually
answers this question.
--
Thomas Roessler, W3C <[email protected]>
On 24 Apr 2009, at 18:02, Scott Wilson wrote:
In our system when a widget is instantiated we generate our own
instance hashes which we append to the widget URL as a parameter,
and our Storage implementation uses this parameter when it needs to
make a request back to our prefs web service to load preferences,
or to set a preference.
I imagine any UA would put a similar mechanism in place in its
Storage implementation to sandbox the preferences for widget
instances; does that need to be specified?
On 24 Apr 2009, at 09:37, Arve Bersvendsen wrote:
On Thu, 23 Apr 2009 20:17:07 +0200, Thomas Roessler <[email protected]>
wrote:
Guido Grassel is reminding me that the HTML5 storage API keys off
origin. Thy means another wrinkle or the uri scheme/origin
discussion.
Note that only the instantiations of storage, through the
localStorage and sessionStorage, are using origin. The storage
interface itself does not, so I do not see any immediate
consequences with regards to preferences or any uri scheme
discussion.
--
Arve Bersvendsen
Opera Software ASA, http://www.opera.com/
