The API Wars

Although the data is fragmentary and the likelihood of anything significant happening soon is impossible to calculate, something tells me FriendFeed has reached the point where it is in the driver’s seat of social media. Whether that means something important in today’s climate of layoffs and hard looks at viral business models is certainly debatable. But FriendFeed has maneuvered itself into the interesting position of replacing Twitter as the front end to the micromessaging platform it created.

As Paul Buchheit proved with his simple Javascript experiment, FriendFeed’s APIs already allow him or anyone to create a simple Twitterish UI in a few hundred lines of code. While this particular take doesn’t do everything he or I’d want it to, it does viscerally prove the underlying point — that FriendFeed can shapeshift into Twitter while Twitter can’t return the favor. If the user can get all the benefits of Twitter plus those of FriendFeed, why use Twitter?

Currently I write to FriendFeed’s engine from a combination of sources, principally the GChat XMPP IM window in Gmail but occasionally via Twhirl and only via TwitterSpy (a delayed Track tool) when responding to an @reply message from someone I don’t follow. In the latter case, my response is sent via Twitter at the first hop, and then aggregated into FriendFeed. In the other more common cases, my comments are sent from FriendFeed to Twitter.

Buchheit’s prototype code is but the latest in a series of examples that suggest Twitter’s stranglehold on its users is weakening. Another is Power Twitter, a Firefox plugin that enhances the Twitter Web UI with a variety of features, most notably one that decrypts shortened URLs into a full text description of the link based on the target page’s title. The tool also unpacks image URLs and places the image inline, something that FriendFeed does with images, podcasts, and video embeds as well.

These small flourishes turn the Twitter Web page into something I’ll check from time to time, as opposed to never with the vanilla site. Twhirl provides persistence, aggregation of @replies, direct messages, and follow updates in a single stream, making it a better solution until recently when FriendFeed added real time support and, just yesterday, an import tool to synchronize Twitter follows as FriendFeed subscribers. All that remains unavailable is Track, though there are poor-man workarounds using Twitter Search RSS feeds imported into FriendFeed rooms.

Most of these new features have been available for the last few months, but Twitter’s decision to rate limit the ability for third party developers to add services related to following and unfollowing has struck a nerve. Twitter’s ongoing rate limiting of its full firehose has kept Track dead in the water except in a crippled form, and in doing so power users have found it necessary to migrate to mass following strategies that make the flow of data impossible to manage.

The services affected by the new rate limits identify spam and marketing bots and attempt to prune them while autofollowing “real” friends. Unfortunately, these services need to ping the Twitter API thousands of times to determine who is “real”, and according to some developers Twitter is unwilling to improve its APIs to make such verbose approaches unnecessary. It could very well be that Twitter doesn’t want third parties to leverage its platform for services it would like to charge for themselves, but whatever the rationale, the net result is an impasse.

To be sure, FriendFeed might have similar problems if the market suddenly shifted over to them in numbers sufficient to test the smaller company’s resources. Yet Google’s open sourcing of Jaiku and its transplant on top of AppEngine suggests that apps could be built that slowly but surely pipe more and more messages to multiple engines. Already FriendFeed is smart about avoiding recursive loops between Twitter and FriendFeed; it won’t take much to establish an entry point that is effectively identical from a Twitter user’s perspective while also polulating several redundant stores with identical material.

The longer Twitter sends mixed messages about its plans for the services people inevitably will find mandatory — Track and filtering — the more attractive these alternate stores will become. A smart front end that manages whatever constructs people find useful in a way that can be configured to each or several UI patterns can gather significant market momentum over the next few months.

What can slow this down? For starters, Twitter can release new versions of some of these services to blunt the opportunity for its competitors. Some postulate a partnership between Twitter and a larger player, but Facebook’s apparent .5 billion stock offer makes it pricey for a Microsoft who classically would rather build than buy. The Jaiku play seems to undervalue Twitter, not to mention the trouble Google is having with real time in general (Feedburner, phone home.) Yahoo? No.

IBM bought Twitter 1.0 with Notes in the Lotus acquisition, and fueled its entry into half of the messaging market around the Y2K marketing opportunity. Now micromessaging offers users a more elastic construct to model digital relationships across decomposing corporate silos and virtual communities. Twitter sits midway between the walled garden approach of Facebook and the evolution of messaging embodied by Gmail and its IM integration.

It’s no small thing that valuable connections these days flow over direct messages. The Twitter social graph is constrained by the difficulty in managing too many follows, and the follow dance leads to cross-follow relationships that business relationships to emerge from the value of the affinities which attract people to each other. As track and filtering return, tools will allow partitioning of micromessages into the atomic datatypes of both email and IM, with the added attribute of dynamic provisioning based on a project by project basis. This is where the money flows, and is made.

FriendFeed is Twitter’s API. It’s also an RSS aggregator — a direct competitor of Google Reader — and forces Google to confront the absurd latencies of Feedburner and YouTube transcoding for display on the iPhone. Today’s rollout of in-line video playback in Gchat suggests Google is interested in keeping users tethered to the Gmail console and its Adsense engine, but FriendFeed’s APIs are available to developers while Gmail Labs remains an internal sandbox. And Buchheit’s previous startup was Gmail.