Optimised Internet apps. vs. RCS vs. multi-headed clients

OCT 11 2010 NEW REPORT AND BLOG POST ON RCS HERE
I'm currently digging into IMS/RCS for what may turn into an "epitaph" research paper. My current spoken and blogged views on it are well-known, but I feel it is worthy of a more weighty and analytical piece.

I'm currently sifting through assorted vendors' websites, GSMA RCS specs, YouTube video demos and so forth.

I'm struck by one very clear question:

On a half-decent phone, why would anyone want to use a multi-headed / aggregated app, hooked into various social networks and messaging services, rather than an optimised one from the underlying Internet service?

For example: The chance that an operator/RCS-mediated "Facebook experience" is ever going to be better than a native app or browser-based "Facebook experience" is surely zero, isn't it? Or am I missing something? Is the operator-based option solely for low-end devices that can't support proper apps?

Surely, the day a web-based service updates its capabilities with something cool and new (say, a "dislike" button, or innovative photo-upload feature), it can update both its browser and app-based interfaces. But it's stuck with whatever the current device client can support for the operator-mediated version.

I can perhaps see the value of importing some operator data and capabilities (eg presence, billing) inside the Facebook app - but I really struggle to see the rationale for doing things vice versa.

The way I see it, social networks become (relatively) more important through two main routes:
1) Viral adoption of standalone clients or web access because of some unique & desirable features to a specific user community
2) Piggybacking on another successful social network as a platform, and then spinning out to standalone once reaching critical mass

So... what is the "vector" for an operator-based social networking service to become widely adopted? Is there a catalyst for "virality"? Is the fundamental desire for that virality being embedded in RCS's design criteria and specifications? At the moment, I see it as being engineering-led, with little regard for basic behavioural psychology.

I'd contend that for innovative mobile applications to become successful, virality is more important than interoperability. (And then there's openness / extensibility, but that's a whole other story).