https://fuchsia.googlesource.com/modular/+/master/docs/modul...
https://fuchsia.googlesource.com/modular/+/master/docs/modul...
I think the doc you linked to doesn't support your idea; it sounds much more like the replacement for activities and intents in Android than what you described. The service providers have no incentive to supply data independent of a full experience (why would Bing let you show maps in your app without agreeing to a ToS? what if the map provider has different streets than the traffic provider?), the app developers have no incentive to do this (if one phone might display Bing maps and another Google in my app, how does that make my life easier?), and the user probably doesn't want this (why am I seeing grainy black and white satellite imagery in one area and high fidelity color imagery in another?).
Think of the headache gluing all that data together. Every map viewer would need to support the data format for every map provider. Every traffic service would need to expose a protocol for querying traffic data on a stretch of road, and you pray that they mostly work the same and know which streets are which. You hope that every point-of-interest provider exposes the same set of fields and level of detail you need to make them useful. And that doesn't begin to touch on the intersection of rendering things well and the data needed to make that work (which streets to show at each zoom level, arrangement of points of interest, knowing which streets are important enough to show traffic for, special cases like when freeways are closed...).
On gluing things together: look at your nearest npm-based software project. It may not be very pretty, but it works in a reasonable way.
The argument isn't very good in my opinion. Combining independent services differs from combining libraries.
If it's just linking, loose coupling, it might work OK (for something like document creation or messaging) but it won't be integrated at a higher semantic level.
That's pretty much standard behavior in the mobile world. Services and activities are made available and the app petitions the OS to access those "lowest common denominator features".
Although you've decided to reffer to it in disdain, it's also a very basic principle in software design.
>geometric increase in bugs
Only if you believe that developing and maintaining specialized programs somehow increases the bug count "geometrically".
I think if a program developed with one version of n components might have x bugs, a program with two versions of n components may have (n^2)x bugs; that is, for every new version of any given component you add to the system, you multiply the number of possible configurations of the program.
Again, if it's all loose coupling, it'll work (but won't be very interesting). But if you have lots of random calendars, different mapping components, different messaging subsystems - it all becomes much, much more complex. Abstractions leak, break down. They have different ordering semantics, different performance, different failure modes, different concurrency support; there's no end to the number of problems introduced by trying to be flexible about what dependencies you plug in.
"And our database will be one table with Int64 possible properties and Variant data for any type, and..."
Honestly that sounds disastrous. Idealistically, sure it’s a nice idea. Practically, it’d be unusable.
They seem to be doing pretty well.
It is time to give up on Windows Phone.
My WP 10 has received more updates than all my Android devices altogether.