Bryce, Last time I tried polling was actually faster than XHRs if you're sending data back and forth at a very high frequency.
On Fri, May 11, 2012 at 1:46 AM, Anis KADRI <[email protected]> wrote: > Android has a direct bridge JS <-> Java through addJavascriptInterface() > which I believe is the fastest solution BUT WebSockets could be an > interesting approach for iOS. > > > On Fri, May 11, 2012 at 1:33 AM, Julien Bouquillon [revolunet] < > [email protected]> wrote: > >> Have you take a look at how linkedin implemented the native to webview >> bridge ? >> >> http://engineering.linkedin.com/mobile/linkedin-ipad-nativeweb-messaging-bridge-and-websockets >> >> They use a websocket which is must faster. This wont work on all androids >> but could improve performance on new devices >> >> Julien >> >> 2012/5/11 Bryce Curtis <[email protected]> >> >> > I believe XHR is faster response on average and lower overhead than >> > polling. I would expect addJavascriptInterface to be faster than >> > prompt. If someone wants to write a test case for each, then we can >> > try it ourselves and provide results to a wiki page to quantify it >> > once and for all (or until the next Android version). >> > >> > My bet is that the optimal case would be to use loadUrl("javascript:") >> > for Java->JS and addJavascriptInterface for JS->Java. Of course the >> > problem with loadUrl("javascript:") is that it steals focus - though I >> > remember reading somewhere that someone implied they found a solution >> > - don't know what it was. >> > >> > On Tue, May 8, 2012 at 4:00 PM, Anis KADRI <[email protected]> >> wrote: >> > > I do however it's not exactly formalized. I just played around with it >> > when >> > > I was working on the couchdb plugin for android and did some testing. >> I >> > can >> > > put something together though. I think the main issue is not if it is >> > > faster or not but rather if it breaks on devices (other than the 2.3 >> > > emulator) which I seriously doubt. >> > > >> > > On Tue, May 8, 2012 at 1:34 PM, Joe Bowser <[email protected]> wrote: >> > > >> > >> Do you have proof that addJavascriptInterface is faster despite the >> fact >> > >> that we're not using WebView.loadUrl? I seriously think we're >> missing a >> > >> large part of the picture if we don't account for that. >> > >> >> > >> On Tue, May 8, 2012 at 1:30 PM, Anis KADRI <[email protected]> >> > wrote: >> > >> >> > >> > > >> > >> > > I wonder how other frameworks get >> > >> > > around this issue? >> > >> > >> > >> > >> > >> > They don't as it's a 2.3 emulator only issue. >> > >> > >> > >> > +1 for adding addJavascriptInterface back it's just 10x faster for >> > >> > everything other than simple operations (like taking a picture). >> > >> > >> > >> > >> > >> > > >> > >> > > On Monday, May 7, 2012, Filip Maj wrote: >> > >> > > >> > >> > > > Something to chalk up on the todo list: test all of the Adobe >> VAN >> > + >> > >> SF >> > >> > > > office Androids with that app you put out. >> > >> > > > >> > >> > > > >We can bring addJavascriptInterface >> > >> > > > >back, but I don't think it's going to magically make things >> > better. >> > >> > > > >> > >> > > > Maybe we could do a benchmark between XHR, prompt and >> > addJSInterace >> > >> and >> > >> > > > see which one has best performance. And this might sound crazy >> but >> > >> > maybe >> > >> > > > even open our methodology and numbers up ;) >> > >> > > > >> > >> > > > >> > >> > > >> > >> > > -- >> > >> > > Simon Mac Donald >> > >> > > http://hi.im/simonmacdonald >> > >> > > >> > >> > >> > >> >> > >> > >
