On 27 May 2012, at 17:49, Anant Narayanan wrote: > On 05/27/2012 05:11 AM, Marcos Caceres wrote: >> On 27/05/2012 12:36, SULLIVAN, BRYAN L wrote: >>> Re "At install time or when I am browsing apps, how does a server know >>> my screen resolution? Or is this restriction imposed on by the user >>> agent?": When browsing apps, the server can easily access the screen >>> and window DOM attributes. >> Right, but that requires some communication that is implicit in the >> spec. I'm trying to figure out what data is leaving my device and going >> to the server, and why (i.e., what is the expected life cycle model). >> There is all sorts of things that are implied going on behind the scenes >> that this spec eludes to (e.g., installation management/sync across >> devices), and it's good to get a sense of how it all comes together. If >> it's not clear in the spec, then I have a hard time seeing how multiple >> user agents will be able behave in an interoperable manner. > > There is no extra data leaving your device. When you visit a store it will > probe for your current device capabilities, and the store, at its discretion, > can decide whether or not to let the user install an app. There is no > enforcement by the User-Agent at install time. > > Synchronization is an interesting problem that we haven't fully tackled > head-on yet. So we might need to add some UA enforcement at sync time as > opposed to install time. > >>> When installing apps, the installer (browser, app manager, etc) can >>> provide a warning to the user that the app is designed for use on >>> larger screens, and may not work properly on this device. >> Sure, but doesn't that lead to the original complaint that certain >> developers don't want their application to install at all for PR reasons? > > In combination with installs_allowed_from, some apps can choose to publish > only on certain stores with which they have an agreement that users won't be > allowed to install apps on devices they weren't designed for. Sure, it would > be easy to bypass this since there is no UA enforcement, but this would be > limited to a fairly small technical crowd.
Two objections: - If its metadata intended for web app *stores* wouldn't it make more sense as part of the metadata for store submission, rather than an API for browser-type UAs? (Once more I'm CCing the web app stores CG...) - If its easy to bypass, why bother with it? (I once wrote a greasemonkey script that let the Chrome Store work on Firefox :) > > -Anant >
