Facebook needed to buy Instagram because it was creating a mobile-first photo-centric social network. Snapchat is interesting to Facebook because it’s creating a mobile-first ephemeral content social network. Twitter was threatening enough to Facebook because of an asymmetric follow graph that Facebook added the equivalent ‘subscribe’ option. And so it goes…

Farhad Manjoo wrote that Facebook’s Poke isn’t problematic as a direct copy of Snapchat but rather because it should have BEEN Snapchat – as in, why didn’t Facebook think of Snapchat? Here’s one daunting passage:

“In order to realize its hundred-billion-dollar dreams, Facebook needs to forever dominate all of the world’s social interactions. Wherever two or more people are communicating—whether it’s by text, video, pictures, or through games or gifts—Facebook needs to be a part of their conversation. It needs to be especially vigilant against usurpers of young people, the vanguard that decides who’ll rule tomorrow’s Internet.”

(of course we’re all going to look pretty silly in hindsight if ephemeral messages is just a feature of chat, not a standalone product. That’s the counter-argument to, it’s a whole new product experience or one where people want to be anonymous and outside of their Facebook graph).

This morning Josh Constine (of TechCrunch) and I were discussing this question on Twitter. Josh covers Facebook for TC and believes they will be able to evolve the way they manage social graphs to accommodate subgroups and other use cases. That is to say, most people won’t feel the need to create separate social graphs outside of Facebook just to satisfy different use cases if those use cases can also be satisfied on Facebook within a subset of their graph. (Josh – let me know if this isn’t what you meant). In fact, Josh wrote his thesis on this and said he’ll put it online.

My inclination is to bet against Josh – I don’t believe Facebook’s position as the ‘one graph to rule them all’ is established. They’ll continue to be successful and useful for quite a long time – and they may even be the largest single graph – but it’s not going to be the only one of consequence. Why?

1. Managing a dynamic graph within a single service is near impossible: For Josh’s statement to ultimately be true, Facebook needs to solve when I want to be public/FOAF/friends/subset of friends; real name/pseudonym; ephemeral/persistent; etc. Wow, how the heck do you do this? Most folks accomplish this today by applying slightly different rules to the groups of people they connect with on different services. Those silos are easy for the average person to understand as they publish and consume.

2. Maybe each generation needs a space to call their own: We’ve never had a social graph last >10 years at scale. Facebook looks like the best chance to do this but we already see questions about their brand attractiveness to teens. Each new group of kids come of age wanting a space they can discover together and call their own. This is DNA, not computer science. It’s not about tech changing (oh, this is Facebook if it was build only for tablets) – it’s about getting to a dry piece of land when you’re 13 years old and being able to plant your own flag. I don’t see how you get beyond the anthropology of this. Maybe it’ll be each new service + Facebook which survive, and as these kids age, Facebook will grow in importance to them, while shedding their adolescent services. Maybe.

3. Sometimes starting from scratch is what people want. In January of this year, Mike Arrington and I had a satisfying back-and-forth about the need to have a “clear all friends” button in Facebook and whether that would actually solve anything. 12 mths later I’m actually a bit closer to his side — even if it doesn’t solve the fundamental design issues with Facebook (and other growth-oriented social services), they’ve got to make pruning your graph easier. “The right group of friends” isn’t a number or a destination. It’s a fluid, abstract concept. I know know know the Facebook team thinks they can solve this with machine learning and Edgerank but I’m not sure science wins here.

Here’s the rub – I think Facebook has always understood that different services would want different graphs. Wasn’t that the whole idea behind Facebook Connect and their platform strategy? Sure, let a thousand flowers bloom but they’d all be within the Facebook garden. Facebook would own the identity system and the underlying data. Well, while Facebook Connect is still a powerful tool, it appears to be less important to developers now than, say, two years ago. Largely because of mobile, there are other identity systems tied to your OS, your application experiences. People seem more willing to create accounts or use something other than Facebook. One way to view Poke isn’t as an innovation failure, but rather the result of Connect not being the glue of the social web? It’s not a Connect Only world

That’s why I gotta believe Facebook is looking at how to improve Connect. Can they make more data available to developers? Can they provide a monetization floor for any developer using Connect via an ad system or even monthly stipend system? Why did much of the Facebook Connect leadership team leave after the IPO – was this a leading or lagging indicator?

There will be another Instagram, another Snapchat. Facebook can’t buy or fast-follow all of the innovations. Can they?

12/29: Some add’l comment from Josh: “Tho I didn’t mean silo’d graphs won’t emerge, but that they’ll only pick away at FB, not usurp it as mass market identity layer”