- From: Marcos Caceres <marcosc@opera.com>
- Date: Mon, 20 Jul 2009 19:33:16 +0200
- To: Scott Wilson <scott.bradley.wilson@gmail.com>
- Cc: public-webapps <public-webapps@w3.org>
On Mon, Jul 20, 2009 at 7:10 PM, Scott Wilson<scott.bradley.wilson@gmail.com> wrote: > On 20 Jul 2009, at 16:29, Marcos Caceres wrote: > >> On 7/20/09 4:32 PM, Scott Wilson wrote: >>> >>> -1 >>> >>> No, we should keep widget.preferences. >>> If a UA wants to, it can simply implement it using: >>> >>> widget.preferences = window.localStorage; >> >> "If the UA wants to" seems kinda bad... I think the spec should say >> something concrete about this. Either they are the same, or they are not. >> >> We can't have on some platforms authors expecting values being >> simultaneously saved to widget.preferences and window.localStorage, and on >> others not. > > A Widget author should be able to trust the implementation of the Widgets > API, regardless of the UA. For conformance, having the UA implement > widget.preferences reliably seems better for Widget authors. Yes, I agree. Your point of reliability makes it clear that separation between the storage mechanisms needs to be clear (I'll at least add a note to the spec to say that window.localStorage and widget.preference are different). > Whether they can expect window.localStorage to work is a separate issue, and > depends on what the UA is; the Widget author shouldn't need to worry about > this as they should be able to develop widgets to run in any compliant UA, > which could eventually be all kinds of devices and services, including ones > like ours where window.localStorage doesn't really make sense. Agree. >>> If it doesn't, the UA can do its own implementation, which we do for >>> example, as for our UA localStorage is not an appropriate >>> implementation, as our users are more likely to interact with the same >>> widget using multiple browsers on multiple machines. >> >> Ok, I guess this means that your implementation of widget.preferences is >> synchronizing multiple clients through some kind of server-sent events or >> polling. > > Yes. > >> >> However, can't you have a some kind of "synchronizer" that listens for >> Storage event and the sends those to the server to propagate to dependent >> clients? > > Far, far less easily and reliably than the reverse case of re-routing > localStorage to preferences (above). > Ok, I'll have to take your word for that. >>> We can also >>> implement widget.preferences. We can't reimplement the user's >>> browser's WebStorage impl (if it has one). >> >> Ok, understood. That makes sense from a "Wookie" perspective. >> >>> For more detailed arguments for keeping preferences, just search back >>> through the list ;-) >>> >>> (We've already argued this one loads of times, Marcos! Why has it come >>> up again?) >> >> I guess it helps me understand the problem and articulate use cases... >> need a "Wookie" reminder to keep me in check :) > > Any time :-D > -- Marcos Caceres http://datadriven.com.au
Received on Monday, 20 July 2009 17:34:24 UTC